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

Подождите немного. Документ загружается.

PETER A

KEMA

AD NEELEMAN

This preference is language-specific. As Scalise (1988) notes, Somali has left-

eaded com

ounds which are inflected

n the second constituent

(

the nonhead

)

This suggests that in Somali Linear Correspondence outranks Input Correspondence.

As in the Bantu languages, there will be

o language variation for those cases fo

which there is a candidate complying with all constraints. With respect to

pluralization of compounds, we know of no languages in which a right-headed

ompound is marked for plural b

a suffix on the left-hand constituent

(

the

onhead), a situation which would involve

ratuitous violations of both Input

orrespondence and Linear Correspondence. Compare Scalise’s (1988) t

polo

ical

chema for inflected com

ounds:

(

48) a. head to the right; inflection to the right (occurs in English, Italian

and Somali

)

b. head to the left; inflection to the left

(

occurs in Italian

)

c. head to the left; inflection to the right (occurs in Somali)

The missin

option is indeed ‘head to the ri

ht; inflection to the left’.

On at least one interpretation of Linear Correspondence and Input

orres

ondence, it seems that the in

ut in

(

42a

)

can in fact be ma

ed without

iolating either. Suppose that the affix is spelled out twice, both on the head and

inearly external to the phonological correspondent of the left-headed compound:

(

)

a. [[

X Y

]

FFIX

]

[[

/x/ /affix/

]

/ /affix/]]

f the mappin

principles in (35) and (40) require that

some

spell-out of the affix

ccupies the relevant position in the morphophonology, (49b) satisfies both of them.

Indeed, such double realizations occur. Scalise (1988) gives the example in (50a,b);

imilarly,

ta ca

bone

marginally allows (50c) as a plural.

(

)

a. mezza notte

iddle nigh

‘the middle of the ni

ht’

zz-

tt-i

middle-

PL ni

art-

n-i

bon

However, given that the pattern in (50) is not the only one attested, there must

also be a mapping principle that is violated by multipl

phonological realization of a

le affix. This is stated in (51) (which is equivalent to No

er’s (1993) Uniqueness

principl

(

)

uantitative Corres

ondenc

No element in the morpho-syntax is spelled out more than once.

ORD

FORMATION IN OPTIMALITY THEORY

311

Quantitative Correspondence is independently

otivated by the simple fact that,

n the absence of conflicting requirements, affixes are not normally spelled out more

han once. Thus, [/read/ /able/ /able/] does not exist alongside [/read/ /able/].

Althou

h the above data are su

estive of an OT treatment, the

also present a

hallen

e to the framework. This is because one and the same morphos

ntactic input

(

RAL) is realized differentl

dependin

on the specific compound it is attached to

(

com

are

a notte

with double plural marking versus carte regalo with

lural

affix only on the head). In order to capture such a pattern, one would have to assume

a different ranking of the constraints for different pluralized compounds. The

possibility that constraint rankings can vary for individual words and morphemes

as been explored in optimality-theoretic implementations o

lexical phonology, but

he disadvantages of the approach are clear. In its extreme form

it would lead to rote

earnin

The problem repeats itself in a more severe form in En

lish. Sub

ect names

derived from

article verbs can surface in th

ee wa

s, each of them violatin

one of

he mapping principles discussed above (see Yip 1978 and Sproat 1985)

(

)

a. truck filler u

b. passer by

c. co

eou

For more discussion on the interpla

between the various mappin

constraints,

ee Ackema & Neeleman

(

2004

)

It does not seem too much to say that one of the core phenomena of morphology,

and perhaps of grammar in general, is that one form can compete with, and hence

block

others. The classical cases o

such competition involve inflectional

orpholo

as re

ulated b

the Elsewhere Principle. In this contribution, we have

ued that there are man

more examples of competition, which differ from the

in t

f th

nat

f th

andidates and the selectin

constraints. It

eems to us that OT is the natural framework within which to explore morphological

ompetition, although, as we have see

there are some recalcitrant data.

Pete

r A

ckema

eeleman

University of Edinburgh University College London

Department of Theoretical and Department of Phonetics and

pplied Lin

uistics Lin

uistics

40 Geor

e Square

ower Stree

dinbur

h EH8 9L

London WC1E 6B

tati

iff

ce be

wee

n th

atterns: the first is more fre

uent than the

second, while the third onl

occurs sporadicall

12 PETER A

KEMA

AD NEELEMAN

e-mail: packema@staffmail.ed.ac.uk e-mail: ad@ling.ucl.ac.u

EFEREN

Ackema

Peter and Neeleman

Ad. 2004.

eyond Morpholog

Oxford: Oxford

versity Press.

Aissen

Judith. 1999. “Markedness and S

bject Choice in Optimality Theory.” Natural Language and

Linguistic Theor

673-711.

Anderson, Stephen R. 1968. West Scandinavian Vowel Systems and the Ordering of Phonological Rules

Ph.D. dissertation

MIT.

aker

Mark C. 1985. “The Mirror Principle and Morphosyntactic Explanation.” Linguistic Inquir

16,

373

–

416.

eard

Robert. 1995

Lexeme Morpheme Base Morpholo

Alban

: SUNY Press.

obal

ik, Jonathan D. 2000. “Adv

bs: The Hierarch

Paradox.”

lot Internationa

4.9/10, 27–8.

onet

Eulalia. 1991.

orpholo

ter S

ntax: Pronominal Clitics in Romanc

. Ph.D. dissertation

MIT.

onet

Eulalia. 1995. “Feature Structure of Romance Clitics.” Natural Language and Linguistic Theor

607-47.

resnan, Joan W. 1999. “Ex

laini

g Morphosyntactic Competition.” I

M. Baltin and C. Collins

(

eds.

)

he Handbook of Contemporary Syntactic Theor

. Oxford: Blackwell

11-44.

homsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.

De Lacy, Paul V. 1999. “Haplology and Correspondence.” In: P. de Lacy and A. Nowak (eds.), Universit

f Massachusetts Occasional Papers in Linguistics 23. Amherst: GLSA.

Grimshaw, Jane. 1997. “The Best Clitic.” In: L. Hae

eman (ed.),

lements o

Grammar

man,

169

196

Halle

Morris and Marantz

Alec P. 1993. “Distribute

Morphology and the Pieces of Inflection.” In: K.

Hale and S. J. Ke

ser (eds.),

he View

rom Buildin

Cambrid

e, MA: MIT Press, 111-176.

Harle

, Heidi and No

er, Rolf. 2003. “Distributed Morpholo

.” In: L. Chen

and R. S

besma (eds.),

he Second GLOT International State-o

-the-Article Boo

. Berlin: Mouton de Gru

ter.

Hoeksema

Jacob. 1984. Categorial Morpholog

Ph.D dissertation, University of Groningen.

Hyman, Larry M. 2003. “Suffix Ordering in Bantu: A Mo

phocentric Approach.” In: G. Booij and J. van

Marle

(

eds.

)

, Yearbook of Morphology 2002

Dordrecht: Kluwer

245-281.

Hyman, Larry M. and Mchombo, Sam A. 1992. “Morphotactic

onstraints in the

hichewa

erb

tem.”

Berkeley Linguistic Societ

18, 3

–

364

Kager, Rene. 1999.

ptimality Theor

Cambridge: Cambridge University Press.

Kiparsk

, Paul. 1973. “‘Elsewhere’ in Phonolo

.” In: S. Anderson and P. Kiparsk

(eds.), A Festschri

or Morris Hall

. New York: Holt

Rinehart and Winston

93-106.

Menn, Lise and MacWhinne

, Brian. 1984. “The Rep

ated Morph Constraint: Towards an Explanation.”

Languag

519-541.

Marantz

Alec P. 1984.

n the Nature o

Grammatical Relations. Cambrid

e, Mass.: MIT Press.

er, Rolf. 1993.

timal Words: Towards a Declarative Theor

Word Formation

Ms.

Princeton

niversity.

Perlmutter

David. 1971.

eep and Surface Constraints in Synta

. New York: Holt

Rinehart & Winston.

Prince, Alan and Smolensky, Paul. 1993.

ptimalit

heor

Ms. Rutgers University / Johns Hopkins

niversity. Published 2004 by Blackwell, Oxford.

Sadock

Jerrold M. 1991.

Autolexical Synta

. Chicago: University of Chicago Press.

Scalise, Sergio. 1988. “The Notion of ‘Head’ in Morphology.” In: G. Booij and J. van Marle (eds.),

Yearbook of Morphology 1988

Dordrecht: Foris

229-245.

Silverstein, Michael. 1976. “Hierarchy of Features and Ergativity.” In: R. Dixon (ed.),

rammatical

Categories in Australian Languages

New Jersey: Humanities Press, 112-171.

Smolensky, Paul. 1995.

n the Internal Structure of the Constraint Component Con of UG

Han

dou

talk

iven at Universit

of Arizona.

Spencer, Andrew. 1991.

orpholo

ical Theor

xford: Blackwell.

Sproat, Richard W. 1985.

n Derivin

the Lexicon. Ph.D. dissertation

MIT.

Trommer. Jochen. 2004. “Hierarch

based Competition and Emergence

f Two-Argument Agreement in

Dumi.” To a

ear in Linguistics.

ORMATION IN OPTIMALITY THEORY

illiams, Edwin. 1997. “Blocking and Anaphora.” Linguistic Inquir

577–628.

illiams

Edwin. 2003. Representation Theor

. Cambridge, Mass.: MIT Press.

oolford, Ellen. 1997. “Four-way Case Systems: Ergative, Nominative, Objective and Accusative.”

Natural Language and Linguistic Theor

181-227.

ip, Moira. 1978. The Inte

rit

of Lexical Nodes. Ms., MIT.

ip, Moira. 1998. “Identit

Avoidance in Phonolo

and Morpholo

.” I

S. Lapointe, D Brentari and

P. Farrell

(

eds.

)

orphology and its Relation to Phonology and Synta

tanford:

wick

, Arnold M. and Pullum, Geoffre

K. 1983. “Cliticization vs. Inflection: En

lish

’

” Lan

502-513.

tekauer, P. and R. Lieber (eds.), Handbook o

Word-Formation, 315—334

2005

ger

inted in the Netherlands.

PRODUCTIVITY: THEORIES

RIE BA

1. INTR

The productivit

of a morpholo

ical process (whether inflectional o

derivational) has to do with how much (or, in the limiting case, whether) it is used in

ati

f f

h ar

t li

n the lexicon. For exam

le, the affix -

in th

ewo

oxen

is not used very often by anyone in the formation of new

words, and is correspondingly felt to be of very low productivity; by contrast, the

ffi

is frequently used in the formation of many plurals by many people, and is

hus felt to be of high productivity. Note that this formulation leaves a great deal

pen. The term ‘lexicon’ can be interpreted as meanin

some concrete dictionar

he presumed mental dictionar

of the individual or the presumed ideal lexicon of

ome fictional speaker. The formulation

en above is in terms of production, and

we might want to ask whether comprehension is just as important. The words ‘not

ery often’ and ‘many’ in the example seem to be extremely vague, and we might

want to know whether or how such notions can be made

recise. In

articular, we

ight want to know whether any word-formation process can ever be said to be

otally unproductive and what maximal productivity consists of. Then we might

want to know

if the lexicon is intended to be the lexicon of the individual

how this

eflects or is reflected by the vocabulary

f the speech communit

All

f th

ese

ood questions, and it must be admitted at

he outset that we do not have very

atisfactory answers to all of them

Part of the reason for this is that the stud

of productivit

as a phenomenon is

elativel

recent. People have known that so

e word-formation

rocesses are used a

ot and that others are not, that some affixes appear to die off, although they also

appear to be resurrected from time to time (see e.g. Wentworth 1941), but no theory

f productivity has been built upon such ob

ati

we w

ant t

ow w

hat

people made of this in earlier years, we have to read between the lines and interpret

what they say. The observation has also be made that lexical word-formation

processes do not appear to be as ‘regular’ (a term we need to take care in

nterpretin

) as, sa

, s

ntactic processes are. Thus lack of productivit

orpholo

is equated to lexical word-formation, productivit

to inflectional

orpholo

. This in turn has been challen

ed. The claim was made overtl

in an

mportant paper by Chomsky (1970), in which the issue was brought to the attention

f modern linguistics. Since then, morphologists have struggled with the notion of

productivity, trying to understand it and r

fin

it an

ode

l it

ithin m

ode

heories. No new orthodoxy has yet emerged from the discussion, so the theories we

are dealing with are still developing, and developing fast. The chances are that by

he time you read this chapter

t will be out of date, in ways that cannot be predicted

RIE BA

at the time of writing. This makes it both a frustrating and an exciting area to deal

ith

s a terminological note, ‘morphological process’ should be understood as

ncluding not only affixation but non-concatenative processes such as ablaut

(

conside

rise

raise

)

, conversion

(

o emp

)

, back-formation

(

seduct

seduction

) and the formation type illustrated by

tittle

tattle

ince affixal

processes are b

far the most common in En

lish derivational word-formation, the

form the bulk of the examples presented here. In principle, thou

h, the same points

old true for non-affixal word-formation. The stud

of productivit

in compoundin

presents further problems which will not be specifically addressed here, since they

are not specifically addressed in the theoretical literature. For some authors

ompounding in English is so productive that it should be treated as syntax rathe

han as word-formation

;

for others co

pounding is split between syntax and word-

formation; for a third group, it is purely word-formation and the same aspects of

productivity apply to compounding as to d

ivation

(

see Bauer 1998 and references

here for some discussion of the wider problem of the status of compounds, and

Kuiper 1999 for an ar

ument that productivit

is not relevant in compoundin

seen

as s

ntax). Compounds will not be overtl

discussed in this presentation, but if

ompounding is simply another means of word-formation on a par with derivation,

all the discussion here should equally apply to compounds.

This presentation begins with a historical approach to ideas about productivity,

going through the ways in which the idea has developed over the last half-century o

o, and considering the different approache

to productivity that have been taken.

2. PRE-

ENERATI

E THE

RIE

F PR

ITY

t is not clear to what extent there were theories about productivit

in the pre-

enerative era. Certainl

productivit

was reco

nised and discussed within

orphology, but the terminology is not always consistent and neither the status no

he precise nature of productivity is made particularly clear. Good statements about

productivity seem to arise particularly in the historical study of English.

Thus we find the following comment in Quirk and Wrenn (1957: 104): “… it is

ften impossible for us to distinguish processes that were active and flourishing

uring the OE period from those which ha

dce

sed

obe

rmati

ve be

lo-Saxons left the continent of Europe…”

Kruisin

a (1932: 22) talks of ‘livin

’ suffixes bein

reco

nisable b

the fact that

are productive, and adds ‘The treatment of

dead

roductiv

ffix

province of historical grammar.’

espersen (1942) notes patterns, including those that are frequently used and

hose that are never used. For example he speaks (1942: 466) of some ad

ectives

t tak

though it may not be clear why not. This is a matter of

productivity, but not called that.

Looking back over earlier discussions of word-formation, Aronoff (1976: 36)

s that an elementar

intuition about productivit

is that it is the same as the

umber of words produced b

the use of a particular morpholo

ical process:

PROD

CTIVITY

THEORIES

productivity equals type frequency (though see now Bauer 2001: 47-51

)

. To some

extent this idea is set up as a straw man by Aronoff (see below), but it is in any case

ard to find linguists or grammarians from this earlier period using the ter

productivit

in this simple wa

espite these comments, there are two fundamentall

pre-

enerative

heoreticians who deserve special mention, Schultink and Zimmer, and their work is

onsidered in the next sections. Although both present their theories following the

publication of Chomsky’s

yntactic Structures (1957), of which they must have

been aware (indeed, Zimmer cites Chomsky overtly and mentions transformations),

heir presentations rely on earlier theoretical models and they are not influenced by

he formalism of generative theories in the way that Lees (1963) was, although he

as writing at about the same time.

3. SCHULTINK

(

1961

)

Schultink’s definition of productivit

has had a profound influence on a whole

eneration of morpholo

ists, particularl

those who have been trained in the

Netherlands. Because of Schultink, Dutch linguists have taken productivity much

ore seriously than linguists of other nation

and have had a mor

ew o

Schultink (1961: 113) defines productivity as follows:

Onder productiviteit als mofolo

isch fenomeen v

staan we dan de voor taal

ebruikers

bestaande mo

eli

kheid door middel van h

t morfolo

isch procédé dat aan de vorm-

betekniscorrespondentie van sommi

e hun bekende woorden ten

rondsla

nopzetteli

k een in principe niet telbaa

aantal nieuwe formaties te formen. [B

productivit

as a morpholo

ical phenom

on we understand the possibilit

e-users, b

means of a morpholo

ical process which underpins a form-meanin

orrespondence in some words they know, to coin, unintentionally, a number of new

formations which is in

rinci

le infinite. [My translation – LB; cf. Booij 1977: 4; Van

Marle 1985: 45.

]

Some interpretation of this ma

be required. The morpholo

ical process is

uivalent to what others call a Word Formation Rule

(

whether that is a rule of

affixation or some other kind of word-formation

)

. The infinite number of

ossible

formations is determined by the fact that the class of bases is in principle open, and

so cannot be pre-determined; in practice the number of possible formations may be

relatively constrained and the number of act

ual formations very small (consider, for

example, derivatives with the prefix

ep-

suc

h a

tep-mother

ep-son

ep-

elationshi

). The controversial part of this definition, at least in the e

es of current

heorists, is the notion that productivit

concerns onl

the coinin

of words which is

one unintentionall

or unconsciousl

. Althou

h it is well-known that speakers

ak th

rmal r

es o

-f

rmati

hen they deliberately coin new words

(e.g. for poetic reasons, see under Bauer below), it is less clear how the analyst is to

ecognize an unintentional as opposed to an intentional coinage, and thus how the

analyst can

udge what is relevant data. Bearing this question in mind, however,

Schultink’s definition can be taken as one which is widely accepted and indicative of

he interpretation that productivit

iven in morpholo

. Amon

man

others,

RIE BA

Marchand

(

1969

)

cites Schultink, and Marchand’s view of productivity is

ompatible with that espoused by Schultink (see Kast

vsky 1999: 33-34).

4. ZIMMER

(

1964

)

immer’s views can be found echoed by later researchers, even those who are

apparently unaware of his work. He assumes that his readers know what productivity

s, and argues that it is a psychologically real phenomenon (Zimmer 1964: 20),

which is more semantic or pra

matic in nature than it is formal (1964: 18). The

followin

brief passa

e from a footnote (1964: 85, fn 2) is of interest in showin

arl

awareness of issues which are curr

t in the earl

twent

-first centur

: “It

appears on the whole that what we might term “item-familiarity” is less important in

eev

ati

non-

derivatives, possibly partly because

non

eof

elatively low frequency…”

The references to a relationship betwee

frequency and productivity are, to be

sure, not original in Zimmer’s work, and ca

be traced much further back

(

see, fo

xample, Schultink 1992). They can nevertheless be seen as sowing seeds for late

work in the area, b

brin

this relationship into the open in a forum where it was

onsidered b

later American (and American-influenced) lin

uists.

immer attempts to relate productivit

to a hierarch

of fundamentall

semantic

estrictions

which

in cumulation

have

he effect of making a coinage unacceptable.

e is not successful in setting up such a system for creating a statement of

acceptability of English negati

e ad

ectives beginning with

he points out, fo

xample (1964: 84, fn 2), that

unwell

ought not to exist given its synonyms

sick

and

ill

Nonetheless, by attempting to formulate

estrictions on productivit

e beco

ne of the first to show how difficult a task this is, how subtle the lin

uistic

ements involved are, and how complex the notion of productivit

is in itself.

. AR

ff’

n t

that th

productivity of an affix is a direct

eflection of the number of times it has been used to create words in the dictionary is

hat ‘it isn’t fair’

(

Aronoff 1976: 36

)

. It isn’

fair

bec

use

it fail

tak

int

ccou

ow many words are possible with a particular affix.

f a

possible word

is an interesting one, and one that has worried

inguists for some time. Aronoff also puts this in context for us.

ust as the simplest goal of a syntax is the

numeration of the class of possible sentences

f a lan

e, so the simplest task of a mo

pholo

, the least we demand of it, is the

numeration of the class of possible word

of the lan

e. The

reatest difference

between the s

ntax and morpholo

with r

pect to his enumeration is that in

derivational morphology there is a distinc

n t

obe

de be

wee

n th

sses of

ossible words and actual words.

(

Aronoff 1976: 17-18

)

This is a particularly good formulation,

aking clear, as it does, the similarities

and differences between morphological and syntactic views of grammar. Aronoff

PROD

CTIVITY

THEORIES

319

(

1976: 19) goes on to state explicitly that (

ust as there are syntactic rules for the

reation of sentences

)

there must be rules for the creation of new words

rules which

e terms ‘Word Formation Rules’ or ‘WFRs’. We might want to argue with the

etails of what Aronoff sa

s in the passa

e cited, but it does not make an

crucial

ifference from the morpholo

ical point of view. Somethin

oo man

cooks

poil the broth could be seen as an actual sentence of our lan

ust as kin

dom

s an actual word

(

both can be attested

n corpora and the like), and something like

ow do you do

might be an impossible but actual sentence, just as

ength i

ossible but actual word

(

neither could be created in the

ese

tat

f th

anguage system, but both can be attested).

t is worth pointing out here that the focus on possible words in morphological

tudy, though not original to Aronoff, is one which has often been ignored in the

ubsequent literature. Much of level-orderin

theor

, for instance, is concerned with

actual words, and frequentl

with impossible actual words (a class which is called

‘lexicalized’ b

other scholars, e.

. Bauer 1983: 42-61).

Returning to Aronoff’s view of productivity, we find he views productivity as a

elationshi

between

ossible words and actual words

(

Aronoff 1976: 36

)

. The

mportant point that he makes about this is that no affix can be said to be absolutely

ore productive than any other affix (or morphological process); productivity has to

be tied to particular base types. This approach is later discussed in terms of limiting

productivit

to specific ‘domains’ (Van Marle 1985). Aronoff here, in common with

ther lin

uists of the period, considers the affix as choosin

the bases to which it

be added. In more recent lin

uistic theories (see e.

. Gie

erich 1999) this is

ften turned round, and linguists have started speaking in terms of bases selecting

he affixes which can occur on them. Williams

(

1981: 250

)

uses the ter

‘potentiation’ in this sense: if a base undergoes some morphological process, it gains

he potential subsequently to undergo some other specific process. Alternatively, we

fin

scuss

out

put constraints (e.g. Plag 1999)

according to which it is the

well-formedness of the final word which is the crucial factor in determining which

processes ma

appl

to the output of other processes. All of these approaches are

s of attemptin

to determine what is or is not a permissible sequence of

orpholo

ical processes, and this is seen as part of what determines the productivit

f certain

rocesses.

hile Aronoff’s view of productivity as the rate at which a particula

orphological process is exploited with a particular base-type seems reasonable as

an idealized definition, in practical terms it is not possible to use it to provide the

ndex of productivity which Aronoff fore

es unless we make a large number of

extra assumptions. For example, although the notion of an ‘actual word’ appears not

unob

ectionable but even theoreticall

necessar

, we have no accurate wa

decidin

at what point a possible word becomes an actual word. Countin

somethin

uch as whether the relevant word appears in a particular dictionary is only a

ubstitute measure for a word having beco

e actual. As Aronoff

(

1976: 37

)

himself

points out, though, the outputs of the most productive morphological processes tend

t t

obe

–

ith

r in

r m

ntal

ionaries, or in paper dictionaries. Thus any

dictionary-based measure is likely to underestimate the productivity of a

ynchronically extremely productive process, while giving a much better measure of

RIE BA

rocess which has ceased to be

roductiv

. Baayen and Lieber (1991: 804) go so

far as to say that any measure bas

d on Aronoff’s notion is ‘a

licable to

nproductive word formation rules only’.

Even the notion of a possible word is slightly hazy in practice. Matthews (1974:

221-2) uses the example of -

ness

ffix

oco

that th

necessarily a definite cut-off between possible and impossible words.

hiteness

ests is clearl

part of En

lish, purpleness is

in his word

‘insecure’ and

entaness is even worse. The implication is that there is more to the productivit

f a particular affix than a particular domain of application. Bauer (2001: 172-177)

onsiders such examples in a small-scale experimental study, and suggests that some

f the difference may be due to factors such as the length and status of the base

(

derived versus monomor

hemic

)

which are, at least in

rinci

le, available in

deciding what is or is not an appropriate base for this type of formation. Since

nobody has ever done a study in which factors such as these are systematically

bserved, we have no way of determining precisely what a possible word may be. If

ese

ffi

nt t

ode

rmin

niquel

a set of bases (which is more o

ess what Matthews su

ests), the situation is even worse.

In a series of articles based on a number of ps

cholin

uistic experiments

(

Aronoff & Schvaneveldt 1978; Aronoff 1980; Anshen & Aronoff 1981; Aronoff

1983), Aronoff and his colleagues show that speakers behave differently with

espect to words containing the same suffix in different contexts. Thus speakers

prefer to form nouns in -

iveness

from ad

ectives ending i

ive

rath

r than n

-ivit

despite models for both in their mental lexicons. When it comes to ad

ectives

ible

however, speakers prefer to form nouns in -ibilit

rath

r than in

-ibleness

ronoff concludes that speakers react to productivit

as a lin

uistic variable, and

hat such productivit

can be different in different domains (determined b

the base

o which the affix is to be attached). Moreover, productivity cannot simply be

educed to a matter of phonological transparency, otherwise

-ness

wou

dbe

referred suffix in all cases. He

es productivity as being related to frequenc

frequent derivatives tending to be the ones created with less productive morphology.

This notion is developed later by Baayen. Aronoff also notes the relative frequency

f the base and the derived form as being something of interest for productivity, a

easure which is later picked up b

(see below).

In another experiment, based on

he Ox

ord En

lish Dictionar

Anshen and

ronoff (1997) show that ver

productive patterns ma

ive rise to forms which are

not lexically listed. They consider the fate of verbs of the form preposition + verb

ode

rn outpla

underanalyze. They show that some forms with the

osition

with

are retained in modern English (

withdraw, withhold

), though the

formation is no longer used in the creation of new verbs. But there are no current

erbs on this pattern using

(

atblow

fas

for example, were used at an

arlier sta

e of En

lish). This is not becau

se suc

ds we

arli

periods of En

lish: the

were common in the fourteenth centur

. But apparentl

when the pattern ceased to be used, all

he verbs created on this pattern also

disappeared. This suggests that the relevant words were not well-established in the

ental lexicon when the

attern ceased to be

roductive, and thus su

orts the clai