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

Подождите немного. Документ загружается.

HELLE LIEBER

(41) [+dynamic ([

voli

], [

[

])] ; [+dynamic ([

], [+dynamic, +IEPS

([

[

], [+Loc

(

[ ]

)

]

)

]

)

, <base>]

ieber assumes the bipartite analysis

f causatives argued for by Dowty (1979),

Levin and Rappaport Hovav (1995), Rappa

port Hovav and Levin (1998), Levin

(

1999); in such an analysis, causatives are broken down into two subevents which

ht be

lossed as ‘x does

’ such that ‘x causes

to become/

o to z’. In Lieber’s

epresentation [+d

namic] represents an eventive function, [+d

namic, +IEPS]

han

e of state or position function, and [+LOC] a locational function. The

ausative

resultative

and ornative -

-if

ede

alt

much as in Plag

(

1999): for the first two, the base is linked to the [+LOC] argument, and the latter to

the [+dynamic, +IEPS] argument. As in Plag, the inchoative -

-if

dealt with by making the causative function optional.

nlike Plag, Lieber assumes

hat the performatives and similatives lie outside of

he core of meaning for this cohort of suffixes, arising as a sense extension from that

ore. Specificall

, she proposes that the similative and performatives arise when the

econd of the two subevents in the semantic re

resentation of these affixes is

dropped, leavin

onl

the first subevent, the one that can be

lossed rou

as ‘x do

y’. So a performative like anthropologiz

can be analyzed as ‘x do anthropology’

imilati

lik

oswelli

‘x

lik

oswe

ll’

ber argues that because

se e

eou

eco

eof

the affixal meaning, forms with these

eanings should occur less frequently than the core causative, resultative, and

rnative meanings. This is certainly the case: fo

performatives and similatives; fo

-i

there are onl

a couple of performatives and

o similatives

;

and fo

ate

ith

The cohort of verb-formin

suffixes in En

lish has also been of interest

because of the relatively int

cate way in which they divide up the range of available

ominal and adjectival bases. Kjellme

(

2001), who looks only at the conditions on

attaching verb-formi

g suffixes to adjectives, notes that -

-if

generally avoid

attaching to complex ad

ectival bases, except for those ending in -

-ic

and -

ian

They are thoroughly averse to attaching to ad

ectives ending in native ad

ective-

forming suffixes like -

ish

-y

an observation which Aronoff and Fuhrhop (2002)

would likel

attribute to their Monosuffix Constraint (see section 2 above, and

auer, this volume

)

here are far more intricate phonolo

ical constraints on the attachment of these

affixes, however, and a long literature tryi

g to explain them. Plag (1999) gives an

excellent history of this debate, citing both derivational analyses (Gussmann 1987,

Schneider 1987, and Kettemann 1988

)

, and con

raint-based analyses (Raffelsiefen

1996). He proposes his own analysis of -ize

as well as the other verbalizing suffixes,

ouched within Optimality Theory. O

un and Sprouse (1999) give a brief

Optimalit

Theoretic anal

sis of -

as well. I will not review the entire debate here

but briefl

summarize what seem to be the main phonolo

ical constraints on the

The feature

[

IEPS

]

stands for ‘Inferable Eventual Position or State’ and desi

nates a path component of

eaning. With the positive value, the feature signals the presence o

a directed

ath. See Lieber and

aayen (1997) for explanation.

ENGLISH WORD

ORMATION PROCESSES

413

attachment of the productive verbalizing suffixes, basing this summary on Plag’s

nt an

dde

tail

ccou

Plag concurs with Raffelsiefen (1996) that -

generally prefers consonant-final

dis

llabic bases with a trochaic stress pattern (e.

standard

). He points out,

owever

that the constraints on the attachm

t of this suffix are far more complex.

Specificall

, he notes that -

can sometimes attach to vowel-final bases, as lon

hey are trochees (so

ossible but memoryiz

is not

)

. Further, stress la

ses

(

that is, sequences of two unstressed syllables) occur in -

forms “only if a

onsonant

recedes -

”(

so federalize is

ossible, but

emoryize is not

)(

1999:

160), and only if the coda and onset of the fi

al syllable of the base are different (so

federalize is possible, but femininiz

is not) (1999: 164). Finally, -

ann

t atta

o a schwa-final stem

(

so patinaiz

is impossible). Plag uses Optimality Theory to

produce an anal

sis which

oes a lon

towards accountin

for these intricate

Further, Pla

sets his account in the context of the rival verb-formin

suffixes

ate

showing how they carve up the existing phonological territory of base

nouns and adjectives rather nicely.

iting Mahn (1971), Gussmann (1987), and

Schneider

(

1987

)

reviou

accounts

he shows that

refers bases with final

tress, either monosyllabic or disyllabic wit

an iambic stress pattern. It will attach

o bases ending in -

if that segment is deleted (

eautif

). Stress shift occurs only

arel

llabi

). Pla

in fact proposes that -

-i

are phonolo

icall

onditioned allomorphs, essentiall

in a relationship of complementar

distribution

with each other

(

1999: 197

)

. The suffix -

ate

on the other hand, behaves ver

much

does, exce

t that -

ate

all

ows

r th

ati

m-final

nant

well as vowels (e.g.,

alluviate

alluvium

(1999: 214). Again, Plag accounts fo

hese facts elegantly in the framework of Optimality Theory.

3.2.4 Adjective-forming suffixes

English has a wide variety of suffixes that create adjectives of various sorts. By

far the ma

ority of them are Latinate in origin. We will treat denominal and deverbal

adjective-forming suffixes separately.

Denominal adjectives

Amon

the suffixes that create ad

ectives from nouns are -

-ed

esque

ish

less

-l

ous

some

and -

(42) -al accidental, global

-ed bearded, blue-eyed

-en wooden

leaden

-esque Kafkaesque, statuesque

-ful shameful

tactful

-ic historic

athletic

-ish wolfish

childish

-less shoeless

treeless

HELLE LIEBER

These vary in productivity from the privative -

less

which is highly productive, to

some

which are quite unproductive in present da

lish. These suffixes

ece

ved

a fair am

f att

nti

n i

the literature. The

have been deplo

in arguments concerning the overall architect

ure of the theory of word formation

(

e.g., Beard 1995, Spencer 1999), and examined with regard both to their semantic

properties (Gorska 1984, Slotkin 1990, Hudson 1975, Ljung 1970, 1976, Beard

1976, 1991,1993, 1995), and their phonological properties, the latter especially

within the theory of Lexical Phonology and Morphology. We start with semantic

issues.

eard

(

1991

)

studies the difference b

tween relational (RAd

) and qualitative

(

QAd

) ad

ectives. Relational ad

ectives are distin

uished b

the fact that the

do not

ccur predicativel

, and cannot be modified with ‘ver

’ or ‘more’: so a

nuclea

physicist

is not (under the most obvious reading) a physicist who is nuclear, nor can

omething be

very nuclear

e nuclea

r. A typical qualitative adjective, on the

ther hand, can occur in those contexts: so in a

hrase like a minty tast

mint

ccur predicatively (the taste is mint

) and can be modified by

Although the point of this article is to show how a class of so-called bracketing

paradoxes can be treated as matters of scope within a semantic analysis of attribute-

ead phrases (e.

uclear ph

sicis

structural lin

uis

), Beard is careful to point

ut that some denominal ad

ectives are sub

ect to both the relational and qualitative

eadin

and that the ‘bracketin

paradox’ can occur with either readin

; so fo

xample, with the qualitative meaning, a criminal lawyer

an be a lawyer who

ommits crimes or a

erson who s

ecializes

n laws which are criminal (they neve

hould have been

assed

)

, the latter bei

g one of the bracketing paradox readings.

ith the relational meaning, a

criminal lawyer

is some one who practices criminal

aw, the predominant bracketing paradox meaning.

The distinction between RAd

s and QAd

s is thus orthogonal to the problem of

those bracketin

paradoxes which arise with denominal ad

ectives. Beard solves this

problem b

appealin

to a decompositional semantic representation. The difference

n readin

(and hence the perception of a bracketin

paradox) arises because the

adjectival meaning can combine with the meaning of the head noun in different

ways; according to Beard (in an analysis prefigured by Marchand (1966)), if an

adjective is qualitative, it combines as a predicate with some semantic feature of its

ominal head; relational ad

ectives, in cont

ast, “after selecting the feature of the

ead noun they wish to compose with, serve as an argument, not predicate, of that

feature”

(

1991: 220

)

Althou

h to m

knowled

e this issue has n

er been addressed in the literature

t is interestin

to consider whether an

denominal ad

ective can be qualitative or

elational, that is, whether there is no affix which is limited to one meaning rathe

han the other. My (perhaps unsystematic) impression is that some affixes favor one

eading or the other. The suffix -

and seems to favor relational meanings

-ly friendly, shapely

-ous monstrous,

oisonous

-some toothsome,

uarrelsome

-y fruity, sandy

ENGLISH WOR

FORMATION PROCESSES

415

(

chitectu

alkaloidal

ancest

diphthonal

), although qualitative readings are

also possible (

iminal

is an example with both). The suffix

allows either reading

fairly easily. Suffixes like

-ed

-esqu

ish

ous

some

and -

on the other hand

seem to me to produce almost exclusivel

qualitative readin

nother issue re

ardin

denominal ad

ectives concerns the interpretation of

ectives formed with the privative suffix -

less

and the suffix or combinin

for

free

oth Gorska

(

1984

)

and Slotkin

(

1990

)

note that derived forms with -

less

frequently have negative connotations, whereas forms with -fre

have a

ositive

onnotation; for example, describing a soft drink as

ugarless suggests that suga

would have been welcome, whereas describing it as

ugarfre

suggests that the

absence of sugar is a positive attribute. Go

ska (1984) provides an analysis within

the framework of Cognitive Linguistics that

derives the negative connotation of –

less

forms from a connection with a metaphor of bodil

possession (one

enerall

does

ant t

obe

less

headless

), whereas more positive connotation of -

stem from a connection to a metaphor of possession over which an actor has control.

manti

ssue

that ha

ece

ved so

att

nti

n i

eco

inati

f th

ffi

–ed

with nouns of alienable or inalienable

ossession. Hudson

(

1975

)

notes

hat th

esee

obe

a r

tri

n th

atta

f -

it is only partly

productive on single nouns (e.g.,

bearded

but

*legged

*eyed

)

, but very productive

n premodified phrases (e.g.

white-bearded

one-legged

blue-eyed

)

. Within the

ransformational framework that Hudson assumes, it is hard to see wh

there would

be a difference in productivit

: either wa

, the

-ed

form is derived from a sentence

the child has e

es/blue e

es. Further

-ed

always designates inalienable

ossession, so

abea

ded man

a red-roofed hous

are good, but

aca

ed man

reen leafed path are not. Ljung (1976) responds to these questions by teasing apart

several different as

ects of inalienable

session, as well as pragmatic constraints

n th

ati

-ed

nouns, for example, that it would be pointless to coin the word

eyed

in a phrase

an eyed child

because all children have eyes, and the word would

herefore not convey useful information. The issue of alienable and inalienable

possession and its relationship to word formation is one which has not been revisited

n recent

ears, but probabl

should be,

n li

ht of recent pro

ress in lexical

manti

cs.

The group of affixes which c

eates qualitative affixes (e.g., -l

-ed

-ful

ish

ous

) has also figured in discussions of the architecture of morphology. Beard

(1976, 1991, 1993, 1995), for example, discusses this cohort of denominal adjective-

forming suffixes as another which supports

is separation hypothesis. He points out

hat many of the relevant suffixes can carry two meanings, namely [X POSSESS Y]

and

[

X SIMILAR Y

]

. So, fo

example, the ad

ective sand

can denote possession in

a phrase like a sand

beach, but likeness in appearance in

and

hair. In En

lish at

east ad

ective-formin

suffixes can

enerall

carr

either meanin

. Beard su

ests

hat the uniform semantics of this group of affixes supports the idea that adjectives

are formed by a single rule of transposition, with particular affixes being matched

ith particular stems in a separate component of the morphology.

eard (1993) complicates this picture somewhat by contrasting the situation in

English with a number of other languages in which the possessional and the

similtudinal ad

ective-forming function can be split between separate affixes. Fo

ROC

HELLE LIEBER

these languages, he suggests that two s

parate rules are needed, each providing a

different semantic representation. He sugg

sts that the situation in English supports

what he calls a dual derivational route

where denominal adjectives can be formed

either by transposition which

erely involves category cha

ge, or by the addition of

manti

mat

rial t

a n

minal

se.

See also Beard and Volpe (this volume).

Spencer (1999) is concerned with relational ad

ectives in English. He argues that

elational ad

ectives are cases of pure transposition, and su

ests that the

be formed

addin

an A ar

ument (for the semantic function of ‘attribute’) which demotes

he R (referential) ar

ument of a base noun.

inally, this cohort of affixes has been of some interest in discussions of

orphophonology. Perhaps because there are so many denominal adjective-forming

uffixes in English, they can be divided relatively neatly into those that are Latinate

(

ous

-esque

)

and those that are native

(

-ful

less

-l

ish

-ed

some

). The former tend to have stress and phonological effects on their bases,

hereas the latter do not. This division figured prominently in the theory of

enerative phonolo

at least from the time of SPE (Chomsk

and Halle 1968), and

formed the basis of the theor

of Lexical Phonolo

and Morpholo

(Sie

el 1974,

llen 1978, Kiparsk

1982, Mohanan 1986). It was of importance as well as in

Selkirk’s (1982) framework, where the former are added by root-level word

structure rules, and the latter by word-level rules.

lthough it is obvious that this cohort of affix

ece

ved

att

nti

ove

r th

years, that attention has been somewhat fragmentary, most often concentrating on a

few affixes at a time. The time seems ripe for a reconsideration of this group in light

f recent theoretical developments.

Deverbal ad

ectives

mong the suffixes forming adjectives from verbs are -able

-ive

and -or

ffi

-ful

occasionally attaches to verb

, but this is not

roductive.

(

43) -able washable, trainable, commendable, operable

-ive persuasive, inductive, decorative, coercive

-ory condemnatory, investigatory, contradictory

fir

f th

ese

ece

ved

t att

nti

ith respect to the phonolo

of -

able

classic stud

is Aronoff’s (1976)

reatment in which he argues that this formative is in fact two separate suffixes.

sing the notation and terminology of SPE (Chomsky and Halle 1968), he assigns

eve

f -

able

a ‘+’ boundary; this is the suffix that attaches to roots, has

phonological and stress effects when attached, and is frequently lexicalized. The

–

able

attaches to words, does not affect stress or segmental phonology, and

gives rise to semantically transparent formations. Aronoff assigns this suffix a ‘#’

boundar

. It seems safe to sa

that this anal

sis is an artefact of the time in which it

as written. It was superceded earl

on b

one in which the suffix could attach

either at the Root or Word level

(

Selkirk 1982

)

, or in an u

date of the framework of

Lexical Phonology and Morphology in which -

able

cou

dbee

ith

r a L

eve

l 1

r a

ENGLISH WORD

ORMATION PROCESSES

evel 2 suffix (see Giegerich 1999: 22). As Giegerich points out, it makes little

at -

able

as two different affixes, given the semantic similarity between

he Level 1 and Level 2 versions

;

rather it

hould be seen as a single affix with dual

evel membership.

Pla

(1999: 79-80) discusses the phonolo

ical restrictions on the attachment of

ther suffixes which derive ad

ectives from verbs: to the extent that

-ful

attaches to

erbs, it favors verbs of one syllable (

hopeful

) or of two syllables with final stress

(regretful

). The suffix

-or

can attach only to Latinate consonant-final verb stems;

ive

is even more restrictive, attaching to Latinate stems which end in /d/, /t/, or /s/.

ith regard to syntax and semantics, the suffix -

able

has also figured

prominently. Williams (1981) proposes that pa

f th

h atta

thi

suffix affects the argument structure of the ba

se ve

bsuc

h that the internal argument

f the verb is externalized

;

so the ob

ect of a verb like

wash

(

He washed the shirt

(

)

becomes the sub

ect or e

ernal ar

ument of the -

able

form

(

his shi

t is washable

)

Roeper and van Hout (1999) ar

ue for a s

ntactic derivation of forms in -

able

The

how first that these derivations have a

sive interpretation, and then argue for a

ati

n in

h -

able

is an adjective in a full AP projection which takes a VP as

ts complement. The VP then undergoes a passive-like operation which dethematizes

he agent and allows the theme to raise to the specifier of the

P. The verb then

aises to ad

oin to the affix. As evidence for this analysis they claim that when forms

n -

able

rth

r n

minaliz

ed w

ith th

esu

ffix -it

the

obe

Burzio’s

eneralization (i.e. that when the a

ent role

f a

liminated

the verb can also

o lon

er assi

n case to its ob

ect). Their evidence for this claim is unclear,

owever, as nouns never assign case and the preposition

is always required as a

ase marker for com

lements of N.

esu

ffix

ive

-or

and -

ant

form relational adjectives. To my knowledge

here have been no special treatments of their syntax or semantics in the literature.

Marchand

(

1969: 316

)

notes that forms in -

ive

are often paralleled by nominals in

ation

xplosive ~ explosion

deco

ative

deco

ation

and so on.

3.2.5

ollectives

final

ffix

ill tr

at h

esu

ffix

age

ery

which

at first glance create abstract nouns on the basis of noun stems. In fact, this pair of

affixes displa

s a complex pattern of pol

sem

which is discussed in detail in Liebe

(

2004). As noted b

Marchand (1969), both of these affixes

ive rise to some forms

which are collective, some which are

lace nouns, and some which mi

ht be

lossed

as ‘behavior characteristic of’ (examples from Lieber 2004: 148):

(

)

-a

e collectives: ba

e, wrecka

e, pounda

e, pluma

place nouns: orphan

e, parsona

e, hermita

behavior characteristic of: brigandage

collectives: peasantr

, tena

ewelr

, machiner

, crocker

place nouns: eatery, brew

ry, nunnery, piggery, fishery

behavior characteristic of: snobbery, prudery, savagery

ROC

HELLE LIEBER

Lieber (2004: 149) argues that the central meaning of these suffixes is the

ollective one: when they attach to singular count nouns like

peasant

jewel

they

hange the quantitative class of those n

‘

hara

f’

eaning follows fairly straightforwardly from this central meaning if we assume

hat “they are formed on a particular type of nominal base – names for types of

people, often derogatory ones – and if we a

ssu

rth

r that th

ose b

sco

o be construed metonymically.” In other words a base like

snob

tak

n a

etonym

ehavior of snobs

and attachment of -er

n a

dds

eaning

‘

all those things that snobs d

’. The ‘place’ meaning arises from a different

ort of sense extension from the collective meaning. It has frequently been pointed

t that th

se e

n in the opposite direction from ‘place’ to

‘collective’ (e.g., Apresjan 1974, Nunberg 1996): in a sentence like

eattle voted

Democ

atic

the

lace name

eattl

is taken to mean the collectivity of people in

Seattle. Lieber (2004) suggests that this common sense extension can go in the

pposite direction as well. Fro

wanner

as a collective of swans, we get the sense

extension ‘place where collective of swans is gathered’. What is interesting about

the polysemy of these suffixes is that t

ey exhibit exactly the same polysemy,

uggesting that these sense extensions are natural ones.

3.3 Conclusion

here are obviousl

man

prefixes and suffixes that I have not treated here. I

ave tried

however

to touch on most of the groups of affixes that have received

att

nti

n in th

ewo

rmati

n lit

rat

eof

the last forty year

ith

bec

use o

their phonological , syntac

ic, or semantic properties.

bviously, there is still a great

f r

ese

h l

ft t

obedo

onversion – at least in analytical languages like English – is the process by

which lexical items chan

e cate

without an

concomitant chan

e in form. It has

so bee

n kn

n in th

lit

rat

-derivation and functional shift

(

cf. Don

1993). In En

lish, conversion is a partic

larl

productive process: nouns frequentl

become verbs

(

Google ~ to google

)

and vice versa

(

to th

ath

Occasionally ad

ectives become verbs as well

(

cool

to cool

). Conversion has been

a sub

ect of much discussion in the literature on word formation both from the

tandpoint of the nature of the morphological process involved and its relationship to

affixation, and from the point of view of its semantics. In fact, one might view

onversion as a sort of battleground over which various theoretical camps have

fou

ht over the

ears. Much of the histor

of conversion anal

ses up to the earl

1990’s can be found in Don (1993), so I will onl

briefl

summarize here.

ENGLISH WORD

FORMATION PROCESSE

419

Marchand (1969) analyzes c

nversion as zero-derivation

that is

the addition of

a phonologically null affix to a stem. His motivation is that the effects of conversion

frequently seem to parallel that of overt affixation (1969: 359):

f we compare such derivatives as

egalize, nationalize, steriliz

ith

vbs

lik

clean

irt

, tid

, we note that the s

ntactic-semantic patt

n in both is the same: the ad

ectives

are transposed into the cat

‘verb’ with the meanin

‘make, render clean, dirt

’ and ‘make, render le

al, nati

nal, sterile’ respectivel

. In the

aliz

-g

roup, the

ontent element is expressed b

the overt morpheme -

hil

in th

clean

-g

roup the

ame content element has no counter

art in phonic expression. As a si

n is a two facet

uistic entit

, we sa

that the derivationa

morpheme is (phonicall

) zero marked in

se o

clean

‘make clean’. We s

eak of

ero-derived deadjectival vbs.

See Pennanen

(

1971

)

for a detailed criti

f th

Mar

han

ew o

f z

ati

llen (1978) is perhaps the earliest with

n the generative tradition to conside

onversion to be the affixation of a phonologically null element; in such an analysis,

the noun

[

ammer] would be changed to a verb by the addition of a phonologically

empt

affix [

[

hammer] Ø]. Her motivation is primaril

phonolo

ical. In order to

explain the level-orderin

properties of certain overtl

derived forms, she must

assume that zero-affixation takes

lace at

evel 2

after the affixation of level 1

affixes. Like Allen, Kiparsky (1982) justifies his zero-affixation analysis of

onversion on the basis of phonological facts, this time facts concerning application

f stress rules within a level-ordered phonology. As Don (1993: 35) points out,

owever, neither Allen’s nor Kiparsky’s analysis depends crucially on the

postulation of an actual zero morpheme. Rather, their level-ordering facts fall out

from an

anal

sis in which conversion is a directional process, with verbs bein

derived from nouns or vice versa (as opposed to an anal

sis which merel

relates

onversion-mates non-directionall

– see below).

Akin to zero-affixation analyses are those which postulate that conversion

nvolves what might be called rebracketing

r category-change with no concomitant

affixation: in such an analysis, the noun

[

ammer] would simply be changed to the

erb

[

ammer] with no apparent structural change. Strauss (1982) argues that this

analysis is superior to the one in which a zero affix is assumed, since it is impossible

n that analysis to prove whether the affix

s in fact a prefix or a suffix. Assuming a

ate

-chan

rule simpl

finesses the issue, as no affixation is assumed.

illiams (1981) advocates a similar anal

sis

n th

iff

rati

For him, items formed via a cate

-chan

rule criticall

lack heads. This then

an be used to explain why the converted verb

grandstand

takes the regular past

nflection

(

grandstanded

(

)

rather than the ablaut form

(

*grandstood

)

. Inflectional

features are only available for Williams via

ercolation from a head. Without a head,

grandstand

ann

t a

ccess

eve

al a

t f

f th

eve

stand

(from which the

noun

stand

was originally derived), and therefore the converted verb must inflect

ularl

Hale and Ke

ser (2002) do not provide a

anal

sis of conversion per se, but

nsofar as the

are interested in the ar

ument-structural properties of converted verbs

shelve

saddle

their analysis is relevant to the debate on the morphological

derivation of conversion verbs. They propose that verbs like

shelve

are derived by a

HELLE LIEBER

rocess of ‘conflation’, which is a s

ecial form of incor

oration

(

Baker 1988

)

“

according to which the phonological matrix

f a complement replaces the empty

atrix of the governing head” (2002: 11). The verb

saddle

wou

dbede

ved

(

)(

2002: 18

)

saddle

first conflates with the empt

prepositional head, and then with

the empt

verb position. Hale and Ke

ser ar

ue that constraints on s

ntactic

ovement

(

like the Head Movement Constraint

)

lain limitations on the

emantics of conversion verbs, for example, that we could not have a verb

above which would mean something like ‘horse-put’ in They horsed the saddl

ovement is only legitimate from a complement position, and

in th

above is in the specifier position of the empty P, it cannot be conflated with the

empty verb.

arious process-oriented frameworks of morpholo

have also considered the

ature of conversion. For example, within the Separation H

pothesis of Beard

(

1995), conversion is simpl

another form of lexical derivation, where lexical

derivation is the process by which the semantic or lexical category of a word is

hanged. Affixation – that is, the addition of overt phonological material -- is

arate from derivation, and one and the

ati

mark

either by overt affixation, as in the case of suffixing -

r -if

ofo

erbs, or by adding nothing at all, as in the case of deriving the verb

hammer

from

he corresponding noun.

Don (1993) attempts somethin

of a s

nthesis between the structure-based

anal

ses of conversion and the process-ba

ed ones. As in

rocess-theories, he

oses that the

rocess of affixation b

split into a morpho-s

ntactic portion and a

orpho-phonological portion. The difference between his theory and that of Beard,

owever, is that the morpho-syntactic portion of affixation is represented by an

abstract structural element

which he re

resents in ca

ital letters; the

rocess which

forms nouns from verbs is represented as NOM, for example. When affixation

ccurs, NOM is added

(

1993: 100

)

(

)

V P

P N

ENGLISH WOR

ORMATION PROCESSES

421

(

)

ALK N

An overt morpho-phonological affix such as

-al

ment

may be mapped onto

(

if, for exam

le, the stem were

ive

amuse

), or nothing may be mapped

nto NOM. In the latter case

as in the case o

walk

in (46), we have conversion.

final type of analysis of conversion is neither structure- nor process-based.

This is the analysis developed in a number of ways in the work of Lieber (1980,

1981, 1992). Lieber provides arguments that conversion is not zero-affixation, no

ndeed an

directional process at all, but rather is a process of relistin

in the

exicon. When nouns become verbs, the

are simpl

reentered in the mental lexicon

as new forms. The process is not derivational at all, in fact, but rather is more akin to

oinage. Although semantically and historically it is frequently clear which form is

riginal and which derived, synchronically

onversion

airs lead se

arate lives. He

easoning for this position is as follows:

f conversion were zero-affixation

would expect a putative zero-affix to behave exactly as a phonologically overt affix

does. Phonologically overt affixes typically determine the gender, morphosyntactic

features, ar

ument structure, and cate

of their derived forms. But converted

forms often displa

a variet

enders, morphos

ntactic classes, or ar

ument

structures. Lieber

(

1981

)

illustrates this with data from German and Old En

lish.

ieber (1992) gives similar evidence from English: whereas verbalizing affixes like

–i

have relatively uniform effects on argument structure, converted verbs display a

far wider range of argument structures than would be expected from a

phonologically overt affix.

separate but not unrelated thread of inquiry concerns the semantics of

onversion from nouns to verbs.

lark and

lark

(

1979

)

remains an excellent and

nsi

htful treatment of the semantics of conversion. The

ue that innovative

n-t

bco

sco

tit

manti

ss o

f it

tin

t fr

ith

tati

nal it

lik

bachelor

or indexical items like

she

The interpretation of

hese items, which they call ‘contextuals’, “depends on the context, especially the

eration of the s

eaker and listener”

(

1979: 783

)

. Clark and Clark

ose a

number of coo

erative

rinci

les akin to Gricean conversational

rinci

les that

speakers use in understanding a newly coined denominal conversion verb like

teapot

(1979: 787):

HE INNOVATIVE DENOMINAL VERB CONVENTION

In usin

an innovative denominal

erb sincerel

, the speaker

ans to denote

(

)

the kind of situation,

(

)

that he has

ood reason to believe (c) that on this occas

on the listener can readil

compute (d)

niquely (e) on the basis of their mutual know

edge (f) in such a say that the parent

noun denotes one role in the situation

and the remainin

surface ar

uments of the

minal

bde

r r

in th

ati

So if two speakers know that their friend has an unfortunate propensity to stroke

peoples’ legs with teapots (Clark and Clark’s example), one can say to the other that