?tekauer Pavol, Lieber Rochelle. Handbook of Word-Formation

Подождите немного. Документ загружается.

OGNITIVE APPROACH TO WOR

FORMATION

259

2. Typically the components differ in d

pendence, with one relatively dependent,

and the other autonomous. In section 3.2 we noted the importance of this

difference (coordinated at both poles) fo

defining the stem-affix distinction.

ases of mutual dependence (each fills a hole in the other) and non-dependence

(

neither fills much of a hole in the other

)

are not uncommon.

3. T

picall

one semantic component is full

schematic for the composite semantic

tructure. This com

onent is the head of the construction

(

section 3.4

)

. There is

othing incoherent, however, in a double-headed or non-headed construction,

and many exist of both types.

. When the head depends (at the semantic pole) on its companion, you have a head-

omplement structure. When the head is autonomous and its companion depends

n it, you have a head-modifier structu

iff

ces be

wee

n th

m ar

atters of de

ree, and plent

of intermediate cases of several kinds exist.

(

See Lan

acker 1987: 277-377, 1991b: 165-183 for further discussion of these

atters.

)

onstituenc

(section 3.8) ma

also be seen as a part of valence. Some theories

efine sub

ect and ob

ect, or head and modifier, in terms of a constituent (‘tree’)

tructure, which therefore must be fixed. CG does not. Semanticall

-based

efinitions of head and modifier are immediately above in (4); subject and object

ave to do with cognitive prominence, not constituency.

4.2 The morpholo

-s

ntax boundar

t is exceedin

difficult, if not impossible, to come u

with a definition of

‘word’ that works in all lan

es. A reasonabl

ood one mi

ht be “a stem (o

ombination of stems

)

lus an

affixes attached to it.”

But as we have seen

the

tem-affix distinction is a matter of degree (section 3.2) and so is the distinction

between affixes

(

or stems, for that matter

)

, and se

arate words. In CG this is not

een as a crucial issue. Elements can be more or less tightly bound to the

phonological core of a word, and whether or not they are taken as tightly enough

bound to be ‘in’ the word, as long as the proper connections are made as they are

ntegrated into the complex structure, a close-enough sem

nti

esu

ill

tain tha

ommunication can take place, and so it does

’t m

h matt

r if it i

iff

nt f

ifferent speakers.

The model presented here utilizes exactl

the same mechanisms and definitions

for syntax as we have seen for morphology. The two domains are not in separate

‘modules’ of the grammar, and nothing very important hangs on whether a

formation is considered to be morphological or phrasal. It can change from one

ategory to the other gradually over time, and speakers are unlikely to be bothered

by the change at all.

lor’s (2002: 171-175) attempt to characterize t

e notion in terms of a word’s phonolo

ical stabilit

phonological integrity, phonological promiscuity, and s

antic coherence, while interesting and

mportant, is, if taken as criterial, problematic e

n for English, and more so for other languages,

particularly agglutinative languages. It is a reasona

ble approach to characterizing a prototype for the

otion, and it may well be that in the natu

re o

f th

that i

all

we c

do.

260

ID T

UGG

iterate English speakers have their s

nsibilities trained by strongly reinforced

onventions for writing some things separately and others con

ected

but even so

here are cases where contradictory conventions occur (e.g.

can not

vs.

cannot

cup full

vs.

cupful

). Hyphens are used as a sort of orthographic halfway/half-

way/half way point between writing as single words or as phrases. Many languages

ave families of phonological patterns that h

p define words. English is fairly

pical in (usuall

) allowin

onl

one primar

stress per word, and some sounds

pronounced in special wa

s at word boundaries. A mechanic’s manual

ontained the remarkable compound noun torsion bar control arm sprin

anchor

ousing

Although this was written as if it were separate words, it is pronounced

with only one primary stress (on the word

pring

for me

)

, and thus conforms to the

pattern for English single words.

Sometimes differences of constituency (section 3.8

)

can affect word boundaries,

and that again is no problem for CG. For instance, what is up

[

the table

or one

peaker can be [upon

]

the table

for another

and communication still work fine.

There is no necessar

hard-and-fast line between what is or is not part of a

iven

Some have thou

ht that specialization of meanin

coincides with stora

e in the

exicon, and accordingly give it great impo

tance in this regard. CG rejects that

assumption, claiming rather that novel items may be specialized from birth, and

any quite predictable, non-specialized forms are entrenched in the lexicon. Also,

here is an enormous inventory of established English phrasal structures that have

pecialized, not-fully-predictab

e meanings. Many of these behave in certain ways,

r in particular circumstances, more like

ords than like phrases. A lar

e class that

ften does this are ‘verb-particle’ structures like throw u

all down

et over

stick

out

and so forth.

In certain cases English conventions allow whole

hrases to be converted into

what are effectively single words. T

aditionally these are often written with

yphens: e.g.

he’s a know-it-all

;

e’s a really off-the-wall kind of gu

with an in-

your-fac

put-up-or-shut-up kind of an attitud

Again, the difference between what

are word-formation patterns and what are phrasal syntactic patterns is not a

and-fast

4.3 Inflection vs. derivation

Some affixes make major changes in their stems, often producing a composite

tructure in a different grammatical class (section 3.1, Figure 3). -

able

-l

(

cf.

Fig. 9) are of that type. Such affixes tend to act unpredictably and not to be highly

productive, though some are quite predictable and productive. Other affixes make

nly relatively minor, predictable changes, and may be required by certain kinds o

onstructions. Some of these are highly productive. An example would be the suffix

PER

LAR

as in

ackle

fir

t kin

n a

derivational affixes

and the second as i

flectional. As usual

the differences

between them are matters of degree, and very many affixes are derivational in some

ects and inflectional in others, or somewhere between inflectional and

OGNITIVE APPROACH TO WORD

FORMATION

261

derivational for a given parameter (

uggy 1985, cf. Bybee 1985: 81-110). (The

plural suffixes in Figures 4-6 are examples.)

There is a strong tendency in many languages for more strongly derivational

affixes to occur next to stems, in effect formin

new, lar

er stems, and fo

nflectional affixes to occur ‘outside’ them. Just as the distinction is a matter of

ree, so this pattern is a tendenc

, and not an absolute rule. The CG account

would certainly not set up separate modules to handle these two types of

orphology.

. WHAT’S SPECIAL ABO

T ENGLISH WORD FORMATION?

The question may be asked, what is there that is special about English word

formation as opposed to word formation in any other language? The answer that CG

ives is a simple one: what makes English special is its particular inventory of

ed wo

dsc

e is true of an

other lan

e, of

cou

se.

ut it is worth pointin

out a few

eneral characteristics of the En

lish set of

established structures which are somewhat unusual and give English words a rather

different ‘flavor’ from those of many other languages.

English has, more clearly than a number of other languages do, an extensive

epertoire of sounds that vaguely suggest

inds of meanings without quite achieving

the separate prominence and clarity of meaning that prototypical morphemes have.

Structures such as

l… an

cl… an

… an

tr… an

…ap

…ash an

…i

…

are of this t

pe. The

are not usuall

productive, but can be so. For instance, a

woman recentl

used the word

lumptuous to describe a full-skirted Za

otec dress,

and the author’s family used the word

lap to mean something like ‘(with) a sharp

on-rebounding blow’. One special type (among quite a few others) involves a

eduplicative syllable or pair of syllables, with a change of the main vowel, often

rom

[

]

to an [

]

or an [

]. Exam

les would be

riff-raff

hilly-shall

fiddle-faddl

kitt

-ca

-za

-slo

-to

-ho

lip-

lop

sin

-son

Kin

Kon

etc.

These and other ‘ideophonic’, ‘phonaesthetic’ or ‘sound-s

mbolic’ structures

(

Rhodes and Lawler, 1981, Hinton et al. 1994) give certain English words a kind of

‘ring’ and a sense of appropriateness to their meanings that I do not know how to

duplicate in any other language I know. (Tho

languages, of course, have special

tructures of their own that cannot be duplicated very well in English.) Under CG

he schemas expressing the commonalities of these structures are perfectly normal

chemas that

ust happen to be limited in the number of their sub-cases and the

percenta

e of the vocabular

the

ht characterize, and tend not to be salient in

eakers’ minds. Other theories often cons

der them ‘peripheral’, and ma

exclude

hem from a language’s word formation ‘component’ (though they may be

accounted for by non-morphological

exical redundancy rules). Under CG they fit

perfectly well into the spectrum of stru

tural types that may be expected.

English exhibits a definite propensity for diminutivity, even monosyllabicity, in

ts lexical formulations; on the other hand, it likes long words too. It is like

‘isolating’ languages in the first quality, a

d resembles ‘agglutinative’ languages in

2 DA

ID T

UGG

he second. This is in some degree a result of English being historically an amalga

f Germanic Anglish and Saxon, which contributed lots of small, pithy words to the

ocabulary, with Latin-based French, which brought in lots of derivational

till a n

cy, 800 years later, for the short words

o be those of Germanic origin and the long ones to come from Latin.

n ubiquitous manifestation of the aforementioned propensity for diminutivity is

he prevalence of

ivation

where

ou simpl

plop a word into a construction

equirin

the ‘wron

’ cate

and let people fi

ure out what

ou mean. As Calvin

(of Calvin and Hobbes) said, “Verbin

noun

weirds lan

e.” And of course as

such coinages catch on and are entrenched and start to form patterns, the language

hanges. Many kinds of zero-derivation are quite common and even productive in

English, including noun-to-verb and

erb-to-noun

or intransitive-to-

ausative/transitive and causative-to-intransitive changes. Examples would include

(

rea

)

wheel

vs.

wheel

(

it ove

)

cook

(

potatoes

(

)

vs.

(

a good

)

cook

the

doo

)

opened

vs.

(

she

)

opened

(

the doo

)

(

she

)

ives

(

the ca

)

vs.

(

the ca

)

ives

(

asil

). One common result is that it comes to be difficult or impossible (and

nnecessar

) to determine for sure which kind of stem an affix is attachin

to. E.

. is

reck-a

ved

m th

r fr

m th

eve

eck

If it is the verb

is it the

ntran

iti

ve ve

r th

tra

itive one? In some languages this kind of indeterminacy

s much less often tolerated: stems tend to stay put in their categories, and affixes

attach to only one kind of stem. The habit of making words out of phrases (section

.2) is also related to the zero-derivation habit, and is somewhat unusual among the

world’s languages.

lish is also unique (thou

h not entirel

unprecedented) in its combination of

a lon

histor

of writin

and a prolific

iterar

tradition, enormous numbers of

speakers and readers, vast

raphical distribution, and usa

e as a

ranca

for

any peoples and professions. It is spok

as a second language by great numbers,

and by many conversant with other cultu

es. It has enjoyed very widespread

diffusion through radio and television, which of course was not an option for any

anguage up till the last century. English speak

rat

ed o

eve

(more than speakers of some other languages) words of non-English origin. It is also

a part of the culture of many English-speakers to en

oy and applaud (and thus

encoura

e) clever and creati

e word-play. All of these f

actors contribute to a great

ichness of the En

lish lexicon.

N: IMPLI

ATI

F A

CCOU

NTIN

R M

RPH

HEMA

Let us summarize some of the implications of anal

zin

word formation patterns

as schemas. A number of these ideas contrast with the ex

lanations other theories

ive in terms of rules or templates of various sorts for combinin

buildin

-bloc

hemes.

• These are ‘bottom-u

’ rather than ‘to

-down’ structures. The primar

data are

he specific cases from which a generaliza

ion is extracted. These are typically

OGNITIVE APPROACH TO WORD

ORMATION

263

stablished as units in their own right, and if the generalization were removed

from the grammar they would not necessarily disappear with it. This coexistence

f structures at different levels of generality (schematicity) is extremely central

o the theor

• Th

es o

r th

eed

t r

emble at all the structures the

are

upposed to produce. This is especiall

true of various sorts of filters

which

decry bad or undesirable patterns rather

than producing good ones. Schemas by

definition embody the commonality of the established (good) patterns, and are

automatically activated in the mind every time a good structure that they describe

is used.

• Structures (at all levels of schematicity) can vary greatly in how well-entrenched

hey are in a given speaker’s mind. Besides

hat

it is to be

pected that speakers

will differ amon

themselves in the de

iff

stablished or conventionalized, as lon

as the inventor

of them matches well

nough for communication to take place. Su

ariati

anguage change.

•

The schemas for words and word-pieces a

e bipolar, symbolic structures like the

particular cases. Morphology and syntax, like lexicon, deal with paired sounds

(or other signals) and meanings. Most of what

we have discussed is as relevant to

yntax as it is to morphology.

• Phonological generalizations or rules are not different in kind from semantic o

orpho-syntactic or morphophonemic rules or generalizations. They are all

chemas, constituted by what is common to the cases they subsume.

•

These generalizations are not expected to be absolute. They cover the cases that

hey cover to the degree that they cov

ce o

f r

ant

ses

which they do not cover, or even of ot

er cases and patterns which contradict

hem, does not den

their existence.

• There is an important difference between small-scale, ‘iff

’

eneralizations like

hose of 6.aa-ac and ae

or the fl… an

…a

flap, and robust, highly

ntrenched schemas supported by thousands of examples, like those of 4.u, 6.ad,

6.af, and 6.ag. Yet the difference is a matter of degree, not an absolute difference

n kind. The mechanisms that are necessary to account for the ‘iffy’ cases

function perfectly well to account for the highly entrenched, productive ones as

well. There is neither reason nor any

otivated way or place to draw a line

between a ‘component’ or ‘module’ for ‘re

ular’ or ‘productive’ morpholo

and one for less-re

ular or less-productive morpholo

•

etworks of schematic relations can become quite complex: it is absolutel

ypical for a structure to elaborate

(

and thus receive

anction from

)

a number of

iff

•

Some structures

(

com

onents

)

sanction

er sub

arts of a structure. Others,

ncluding constructional schemas, sancti

n th

iff

ce be

wee

he two types is one of degree.

•

The same sanctioning structures, be they

constructional schemas or components,

are used both b

speakers and b

hearers, for construction of novel structures and

ID T

UGG

e-construction of already-established ones, and for analysis of novel or

ed o

es.

•

Structures differ in analyzability, the extent to which one can discern in the

sanction from components. The difference between components and non-

omponents is a matter of degree.

• Stems sanction a subpart of a complex structure. Affixes are strongly linked to

onstructions which sanction the whole but leave the affix’s com

anion

(

the

stem

)

uns

ecified. The difference betwee

affix

a matt

ree.

•

Process morphemes, zero morphemes, reduplications and other ‘exotic’

orphological mechanisms can be handled with the same apparatus as the more

ommon additive morphemes (‘normal’ roots and affixes), and differ from the

nly in degree, not in kind.

• Schemas can sanction novel structures, functionin

as patterns for thei

formation and thus legitimizing them. Such sanction can range from being so

ompletely automatic that one hardly notices it (e.g. 4.u

4.v) to norm-bending

ses

lik

Freddag

flumptuous

•

Productivity, the habit of using a schema to sanction novel structures, is a matte

f degree.

•

onstituency, the order in which pieces are put together to form, or analyzed out

f, a complex structure, is often variable, with the same overall meanin

and

sound achievable throu

h alternate orders.

• The difference between what is and what is not part of a

iven word is not

always clear, and may vary.

•

s a limiting case there may be absolute rules or laws of morphology in a given

anguage, but they are very much the exceptional case. Most categories and

distinctions are matters of degree.

avid Tu

ggy

ummer Institute o

Lin

uistics

Apdo 2206

4000 M

xico

D.F.

México

-mail: david_tuggy@sil.org

REFEREN

ybee, Joan. 1985.

orphology: A study of the relation between meaning and form. Typological Studies

n Languag

Vol. 9. Amsterdam and Philadelphia: John Benjamins.

ybee, Joan. 2001.

honology and natural language us

Cambridge Studies in Linguistics

94.

ambridge: Cambridge University Press.

illmore, Charles J.; Ka

, Pau

; and O’Connor, Mar

. 1988. “R

ularit

and idiomaticit

rammatical constructions: the case of

let alone

” Language

64,

-5

OGNITIVE APPROACH TO WOR

ORMATION

265

oldberg, Adele E. 1995. Constructions

A construction grammar approach to argument structure.

hicago: University of Chicago Press.

inton

Leanne

;

Nichols

Johanna

;

and Ohala, John J.

(

eds.

)

1994.

ound symbolism

Cambridge:

ambridge University Press.

Lakoff, Geor

e and Johnson, Mark. 1980.

etaphors we live b

Chica

o: Universit

of Chica

o Press.

Lan

acker, Ronald W. 1987. Foundations o

nitive

rammar: V

l. I, Theoretical prerequisites.

tanford

tanford Universit

Press.

Langacker, Ronald W. 1991a. Foundations of Cognitive grammar: Vol. II, Descri

tive a

lication.

tanford

tanford University Press.

Langacker, Ronald W. 1991b.

oncept, image, and symbol: the cognitive basis of grammar

Berlin

ew York: Mouton de Gruyter.

Langacker, Ronald W. 2000.

rammar and conceptualization. Berlin and New York: Mouton de Gruyter.

Moore, Terence, and Carling, Christine. 1982. Language understanding: Towa

ds a post-Chomskyan

inguistics. New York:

t. Martin’s Press.

osch, Eleanor. 1973. “On the internal structure of p

ceptual and semantic categories.” In: T. E. Moore

(

ed.

)

, Cognitive development and the acquisition of language

111-144. New York: Academic Press.

edd

, Michael J. 1979. “The Conduit metaphor – A

ase of frame conflict in our lan

e about

e.” In: A. Orton

(ed.),

etaphor and thou

, 284-324. Cambrid

e: Cambrid

e Universit

ess.

hodes, Richard A., and Lawler, John M. 1981. “Athematic meta

hors.”

CLS

17, 318-342.

Sereno, Joan A., and Jongman, Allard. 1997. “P

ocessing of English inflectional morphology.”

emor

and Cognition 25

425-37.

Taylor, John R. 2002. Cognitive grammar

Oxford: Oxford

versity Press.

Tuggy, David. 1985. “The inflecti

nal

/de

ati

nal

tin

” Workpapers of the Summer Institute o

Linguistics at UN

209-222. Grand Forks

ND: SIL.

Tuggy, David.1992. “The affix-stem distinction; A Cognitive gramma

analysis of data from Orizaba

atl

”

ognitive Linguistics 3/3

237-300.

267

tekauer, P. and R. Lieber (eds.), Handbook of Word-Formation, 267—284.

2005 Sprin

er. Printed in the Netherlands.

WORD-FORMATION IN NATURAL MORPHOLOGY

. DRE

LER

1. INTR

hereas ‘natural’ has often been used b

lin

uists in an inductive wa

as a

synonym of intuitively plausible or of

cross-linguistically frequent, in

Natural

orpholog

(

henceforth NM

)

, ‘natural’ is synonymous with cognitively simple,

asily accessible (esp. to children), elementa

y and therefore universally preferred,

.e. derivable from human nature

or with the terms unmarked or rather less marked:

for in NM it is clearly a relative, gradual concept. This establishes a first difference

with its two main forerunners, namely Jakobson’s (1941) concepts of markedness

(includin

his introduction of the Peircea

semiotic notion of iconicit

) and

Stampe’s (e.g. 1973) model of Natural Phonology (cf. Wurzel 1988a).

Properties which go back to these forerunners are the search for a motivation of

orphological principles and preferences outsides morphology, both a) in othe

omponents of language such as syntax, phonology, discourse and the lexicon and b)

utsides the linguistic system, i.e. psyc

ological, neurological,

ognitive, semiotic

otivation; the assignment of equal importance to internal and ‘external’ evidence

(particularly from diachronic change and language acquisition); functional

xplanation (cf. Dressler 1995, 2002).

Functional explanation starts with the two main functions of language, the

ommunicative and the cognitive function, i.

. the function of supporting cognition.

ord formation serves both main functions. The specific main functions of wor

formation are partiall

shared with inflectional morpholo

, i.e. semantic and formal

otivation of a complex or derived word b

its parts or, in a rule format, b

its

base

(

)

and the word-formation rules d

iving the complex word. Only word

formation, but not inflectional morphology, has the lexical function of lexical

nrichment. In addition to these ma

or functions, there exist very minor functions,

uch as the textual function of providing anaphoric, cohesive links between chunks

The theory of NM (cf. Kilani-Schoch 1988, Dressler et al. 1987, Dressler 2000a)

nat

raln

ess

cove

r t

rm f

r a

of more s

ecific terms which are defined

n three subtheories: 1. a universal markedness theor

of s

stem-independent

orpholo

ical naturalness (cf. Ma

erthaler 1981), focussin

on universal

preferences, 2. a theory of typological adequacy (cf. Dressler 1985a, 1988a), 3. a

heory of system-dependent naturalness or system-adequacy (cf. Wurzel 1984,

Dressler and Ladányi 2000). These subtheories function as subsequent filters on

possible and probable words of a language: what is allowed or even preferred

niversally, may be rendered dispreferr

d or even disallowed by typological

adequacy and then by language-specific system adequacy

268

LER

2. UNIVERSAL

SYSTEM-INDEPENDENT MORPHOLOGICAL

RALNE

.1 Preferences

his subtheory of universal markedness is a preference theor

(

cf. Vennemann

1983, Dressler 1999

)

. Preferences of NM differ from violable constraints of

Optimality Theory particularl

insofar as they are based on external foundations, i.e.

ot postulated ad hoc, and as their interaction is not extrinsicall

ordered but

depends both on principles of interaction (cf. section 4.3) and on the filterin

the

ubsequent subtheories (see above). Much

scuss

f th

ese d

iff

ces c

found in Dziubalska-Kolaczyk

(

2001

)

. More similarities to NM can be found in

ognitive Linguistics (cf. Croft and

ruse 2004

)

. For a com

arison between

ognitive Linguistics and Optimality Theo

y, see Nathan & Winters (2001).

niversal preferences hold both fo

rammatical and extragrammatical

orphology (often called plain vs. expressive morphology, cf. Zwicky & Pullu

1987, Dressler 2000b). What unites extragrammatical morphology are various

iolations of universal properties. Thus echo-words such as

tick

tock

ip-ho

zag

be formed either from a left-hand or a ri

ht-hand base or have no existin

base at all. Also blends

(

contaminations

)

such as

smog

(

fro

smoke

)

and

ther abbreviatory operati

ns are extragrammatical.

aturalness does not refer to any global or overall preference, but to what is

universally preferred separately on each

f a r

tri

f nat

raln

ess

parameters established deductively by universal markedness theory. Each of the

following parameters and its preference degrees are deduced from the extralinguistic

bases mentioned in 1

(

cf. Dressler et al. 1987, Kilani-Schoch 1988, Dressler 2000a

)

.2 Preference for iconicity

he best-known semiotically-derived parameter is the parameter of

conicit

based on Peirce’s (1965) concept of icons with its hypoicons, i.e

mages

diagrams

and metaphors.

mages are the most iconic hypoicons insofa

as they represent a direct similarity

between signans and signatum. A case in point is the preference for expressing

diminutive formation via palatalit

(cf.

ressler & Merlini Barbaresi 1994

2001

pace Bauer 1996). This is clearest with small children. Thus

in a famous

diminutive-formation experiment devised b

Berko (1958), the

oun

est a

roup

formed a diminutive from a nonsense word

such as

by changing the vowel into

the most palatal one and thus producing wig. Older children

roduced suffixed

diminutives with the

alatal suffix, as in wugg-i

whereas adults

roduced less

ico

imin

ves suc

h a

wug-le

wug-e

tte

swe

Diagrams represent an analogy between the relations of the signans and the

ignatum. Thus the fact that in the compound

stone

eseco

stone

ORMATION IN NAT

RAL MORPHOLOGY

269

s both the morphosemantic and the morphotactic head, whereas the first element

ner

is the non-head on both levels (i.e. signans and signatum, and the same holds,

with reversed roles, for its twin compound

stone co

r) is diagrammatic. For it

that a

stone

is a t

pe of

stone

with which it shares its s

ntactic and

semantic features of bein

an inanimate, concrete noun (morphosemantic headhood)

and that the plural suffix is affixed to the same second element (morphotactic

eadhood), and this establishes an exact, diagrammatic parallel between the

orphotactic and the morphosemantic head-nonhead relations in the signatum and

he signans. (Semantic headedness is most evident in so-called pleonastic

ompounds, where the head is a hyperonym of the non-head, as in

oak

(

loomer 1996). Such optimal diagrammaticity holds for most English compounds

and for all the productive ones, but not for the type pickpocke

where in the

natum the first element

overns the sec

nd one, whereas the plural suffix is still

added to the second element

(

cf. also below section 4

)

The most important instance of morpholo

ical iconicit

occurs in the

arameter of

constructional iconicity

(cf. Mayerthaler 1981). According to

Peirce’s (1965) classification of icons, the following examples of English

derivational morphology can be classified as follows: the noun

denominat-or

ved

m th

eve

denominate

in a diagrammatic way, because of the analogy of

addition in meaning and form, whereas the derivation of

s only

etaphoric (i.e. with weaker iconicit

), because addition of meanin

is paralleled

not b

addition but modification of form. Finall

, the conversion of

to cut

acut

appears to be non-iconic, because addition

f meanin

is not paralleled at all b

hange in form. Crocco-Galèas, however,

as maintained in several

ublications

(e.g. 1990, 2003a) that conversion represents a morphological metaphor, i.e. a

etaphorical operation which is signalled by the difference in syntagmatic

ollocation in different syntactic frames.

The anti-iconic operation of grammatical subtraction does not occur at all in

English word formation.

Since the amount of naturalness decreases on this subparameter of constructional

conicit

from affixation over modification,

hen conversion to subtraction

we can

predict that crosslin

uisticall

affixa

ion should be more frequent and productive

han modification and much more so than conversion, not to s

eak of subtraction.

This is true (cf. Dressler 1982) and holds also for English, with the exception of

onversion being more frequent than modification. Moreover conversion is very

productive in English, modification not at all. This makes Crocco’s proposal of

defining conversion a

metaphoric instead of non-iconi

, quite attractive, because

n m

ifi

ati

n an

dco

wou

be both metaphoric and no rank order of

frequenc

and productivit

would be predicted.

Even a hi

her amount of iconicit

can be predicted for extra

rammatical

orphology, since there universal preferences are not curbed by grammatical syste

adequacy. Thus echo words, such as

igzag are highly iconic: the repetition of the

nantal fram

z_g

diagrams repetitions in the word’s meaning, the change of the

owel symbolizes metaphorically change of direction.

xtragrammatical subtraction occurs in all abbreviatory devices and derives

here from a conscious action of economizing. The relatively high amount of