?tekauer Pavol, Lieber Rochelle. Handbook of Word-Formation
Подождите немного. Документ загружается.
PROD
U
CTIVITY
:
THEORIES
331
c
oncerned with degrees of profitability rather than with the question of availability,
although he treats availability as the ex
t
reme end of a scale of profitability.
A
nother thread which runs through discussions of productivity is the importance
o
f limitations on productivit
y
. The existence of principled
rest
r
ictions on
productivit
y
is acknowled
g
ed in Zimmer (1964), althou
g
h Zimmer leaves man
y
r
estrictions unformulated. Booi
j
(1977: 5) however, states clearl
y
that “the de
g
ree of
productivity of a WF-rule can be seen as inversely proportional to the amount of
c
om
p
etence restrictions on that WF-rule.”
T
he existence of restrictions on
p
roduc
t
ivity is not questioned (although how far
t
hese are inevitably matters of competence is considered in Bauer 2001, and the
v
alidity of Booi
j
’s equation above is queried for inflectional morphology by Dressle
r
1997). Statements such as, for example, that English nominal -
al
is added only to
l
v
erbal bases which are stressed finall
y
(
a
rr
ival
,
re
f
erral
;
bu
r
ial
is etymologically
l
unrelated) are commonplace in the literature;
s
ee, for example, Bauer (1983: 84-99,
1994: 3356 and sources cited there
)
. One matter treated in Bauer
(
1983
)
as a
r
estriction on productivity is the question of
b
locking
,
which Kiparsky attempts to
build in to Lexical Morphology as a matter of rule-ordering, although that is almost
c
ertainly too restrictive (Bauer, forthcoming). Further, the domains in which word-
formation rules can apply (see Aronoff, Van Marle, Plag) can be seen as some of
t
hese competence restrictions. While not underestimating the importance of such
r
estrictions, Bauer (1992, 2001) does suggest
t
hat competence restrictions cannot be
the only things influencing apparent profitability. The fact
that productivity can vary
t
dependin
g
on social environment and st
y
le (Pla
g
et al. 1999) also seems to indicate
that f
o
rmal
co
n
s
traint
s
ar
e
n
o
t all that i
s
in
vo
l
ved.
A
distinction between creativity and productivity has been discussed here mainly
i
n reference to Bauer
(
2001
)
, but simil
a
r
notions are already there in Schultink
(
1961), Booi
j
(1977), Van Marle (1983) and, as Booi
j
points out, in Chomsky’s
(
1964: 22
)
r
ule-governed creativit
y
an
d
r
ule-changing creativit
y
.
A
ll this can be taken at a slightly more abstract level
,
and we can ask how much
o
f productivit
y
is a matter of competence and how much a matter of performance (o
r
h
ow much is a matter of
g
rammar and how much is a matter of use). We mi
g
ht sa
y
that an
y
one who has tried to specif
y
an
y
Word Formation Rule in an
y
detail (i.e.
be
y
ond a basic schema of the V + er
→
N
type) is attempting to account for the
N
l
imited productivit
y
of that rule. But such approaches assume that limited
productivity is part of our competence, that is, it is part of the grammar of English.
Many authors explicitly or implicitly see productivity of morphological processes as
being nothing to do with the grammatical system, but rather an epiphenomenon of
t
he historical exploitation of the grammatical system. This can be linked to the
earlier question of whether productivity is or is not a matter of degree. For example,
i
n Štekauer’s onomasiolo
g
ical theor
y
of word-formation, there are a number of
possible different wa
y
s of formin
g
a suitable namin
g
unit to correspond to a
g
iven
n
eed, and one happens to be chosen. The r
e
s
ult, accordin
g
to Štekauer (e.
g
. 2000: 3)
i
s that all word-formation rules are fully productive, they may simply not be used on
a given occasion. This is a view of productivity as availabilit
y
.
While this is a
perfectly coherent point of
view, it simply ignores the question of profitability.
f
Presumably, within this approach profitability would not be seen as anything to do
33
2 LA
U
RIE BA
U
ER
with the com
p
etence of s
p
e
ak
e
r
s.
Th
e
alt
e
rnati
ve v
iew would ask how s
p
eakers
c
hoose between the alternatives on any gi
v
en occasion, and thus why some word-
formation rules appear to be given priority at some periods of history and not at
o
thers. It may be that this is entirely a social phenomenon unaffected by linguistic
factors, but there is a lot of evidence that linguistic factors are often more important
than they appear on the surf
ace in such instances. The pr
f
f
ecise mixture of lin
g
uistic
and non-lin
g
uistic factors has not been established or even widel
y
discussed.
A
nother thread is the notion that productivit
y
indicates probabilit
y
.
Thi
swo
rk i
s
c
learly aimed at the profitability side of productivity. We have seen this most clearly
i
n the work of Baayen, but the idea goes back a lot further, at least to Harris (1951:
374-5
)(
see also Aronoff 1983: 163fn, ‘In word formation, the code word fo
r
probability is productivity
’
). Frequency also ties into probability, in that the more
frequent a particular affix is, the more likely we are to meet a type containing that
affix which has not occurred previously in a particular corpus (whether that corpus is
an electronic entity, a dictionary, or a lif
e experience). Thus it is not necessarily
f
f
clear to what extent talking of productivity
in terms of probability is a different
y
approach from talkin
g
about in terms of the number of outputs formed, or, indeed, in
t
erms of the possibility of creating new
o
ut
p
uts
(
see Rainer 1987
)
. The focus is no
doubt different; the phenomenon may well be the same. Some authors, though, like
D
ressler and Ladányi (2000), explictly deny the feasibility of deriving measures of
productivity from frequency data.
It should be noted that nearly all the work on morphological productivity has
been carried out within formal approaches. Work based on a separationist view of
m
orpholo
gy
such as that espoused b
y
Beard (e.
g
. 1998) has not made a deep
i
mpression on studies of productivit
y
, althou
g
h some of Pla
g
’s (1999) work can be
v
iewed in this light. Again Dressler
a
n
d Ladányi (2000) explicitly rule out any
s
eparationist account of productivity. Accordingly, despite a call for this in
Kastovsky (1986), we have few studies
o
f productivity which consider how, fo
r
e
xample, ad
j
ectivalizations of verbs are formed and relatively many which conside
r
h
ow words using the suffix -
ee
ar
e
f
o
rm
ed o
r h
ow co
n
ve
r
s
i
o
n
wo
rk
s.
It i
s
n
o
t
c
l
e
a
r
h
ow much of a difference this makes, although Plag (1999) suggests that the
formation of verbs is rather more re
g
ular than mi
g
ht be supposed if individual
affixes were considered in isolation
,
and Bauer (2001) makes a similar su
gg
estion
c
oncernin
g
nominalizations of verbs.
15. CONCL
U
SION
W
hat perhaps stands out most clearly in all of this is the way in which
m
orphological productivity is being taken mo
r
e seriously as a part of linguistic
t
heorizing than it used to be. Productivity may an epiphenomenon of something
m
ore basic (as in Natural Morphology and i
n
Hay’s theory), it may be two separate
phenomena, availability and profitability, it may be a matter of grammar or it may
not, but it is no lon
g
er bein
g
i
g
nored. This is pro
g
ress.
PR
O
D
UC
TI
V
ITY
:
THE
O
RIE
S
333
L
au
r
ie
B
aue
r
S
chool of Linguistics and Applied Language Studies
Victoria University of Wellington
P
O
Box 600
Wellington
New Zealand
e
-mail: laurie.bauer
@
vuw.ac.nz
REFEREN
C
E
S
An
de
r
so
n Kar
e
n Eliza
be
th
.
2000
.
P
roductivity in English nominal and adjectival derivation, 1100-2000
.
U
npublished PhD dissertation, University of Western Australia.
Anshen, Frank and Aronoff, Mark. 1981. “Morphologi
c
al productivity and phonological transparency.”
C
anadian Journal of Linguistics/
R
e
vue Canadienne de Linguistiqu
e
26, 63
-7
2
.
Anshen, Frank and Aronoff, Mark. 1997. “Morphology in real time.” In: G. Booij and J. van Marle (eds.),
Y
earbook o
f
Morpholo
gy
199
6
.
Dordrecht: Kluwer
,
9-12.
A
ronoff
,
Mark. 1976.
W
ord Formation in
G
enerative
G
rammar
.
Cambrid
g
e, Mass.: MIT Press.
A
ronoff, Mark. 1980. “The relevance of productivit
y
in a s
y
nchronic description of word formation.” In:
J
. Fisiak
(
ed.
)
H
istorical Morpholo
gy
.
The Ha
g
ue: Mouton, 71-82.
Aronoff, Mark. 1983. “Potential Words, Actual Words, Productivity and Frequency.” Preprints of the
P
lenary Session Papers of the XIIIth
I
nternational Congress of Linguists
.
Tokyo, 163-170.
Aronoff, Mark and Schvaneveldt, Ro
g
er
.
1978. “Testin
g
morpholo
g
ical h
y
potheses.” Annals o
f
the New
Y
ork Academy of Science 318, 106-114.
B
aayen Harald. 1989.
A
Corpus-Based Approach to Morphological Productivit
y
.
Doctoral dissertation,
Free University of Amsterdam.
B
aayen, Harald 1992. “Quantitative aspects of morphol
o
gical productivity.” In: G.
B
ooi
j
and J. van Marle
(
eds.
)
, Yearbook of Morphology 1991
.
Dordrecht: Kluwer
,
109-149.
B
aayen, Harald. 1993. “On frequency, transparency and productivity.” In: G. Booij and J. van Marle
(
eds.
)
, Yearbook o
f
Morpholo
gy
1992
. Dordrecht: Kluwer
,
181-208.
B
aa
y
en, Harald and Lieber, Rochelle. 1991. “Productivit
y
and En
g
lish derivation: a corpus-based stud
y
.”
Lin
g
uistics 29
,
801-843.
B
aayen, Harald and Renouf, Antoinette. 1996. “Chronicling
The
T
imes
:
p
roductive lexical innovations in
an English newspaper.” Language 72
,
69-96.
B
arnhart
,
Robert K.
;
Steinmetz
,
Sol
;
and Barnhart
,
Clarence L. 1990.
T
hird Barnhart Dictionary of New
E
nglish
.
n.p.: H. W. Wilson.
B
auer
,
Laurie. 1983.
E
nglish Word-formation
.
Cambridge: Cambridge University Press.
Bauer, Laurie. 1992. “Scalar productivity and -
l
il
y
adverbs.” In: G. Booij and J. van Marle (eds.),
Y
earbook o
f
Morpholo
gy
1991
.
Dordrecht: Kluwer
,
185-91.
B
auer, Laurie. 1994. “Productivity.” In: R.E. Asher (ed.),
T
he Encyclopedia of Language and Linguistics
,
Oxford: Pergamon, 3354-7.
B
auer, Laurie. 1998. “When is a se
q
ue
n
c
e of two nouns a compound in En
g
lish?”
E
n
g
lish Lan
g
ua
g
e and
Lin
g
uistics 2
,
65-86.
B
auer
,
Laurie 2001.
M
orpholo
g
ical Productivit
y
.
Cambrid
g
e: Cambrid
g
e Universit
y
Press.
Bauer, Laurie. (forthcoming). “Com
petition in word-formation.” In: A.
m
m
v
an Kemenade and B. Los
(
eds.
)
,
T
he Handbook of the History of English
.
Oxford and Malden
,
Mass.: Blackwell.
B
auer
,
Laurie and Renouf
,
Antoi
n
e
tte. 2001. „A corpus-based study of compounding in English.” Journal
o
f English Linguistics 29
,
101-123.
B
eard, Robert. 1998. “Derivation.” In: A. Spencer and A. M. Zwicky (eds.), The Handbook o
f
M
orpholog
y
.
Oxford and Malden
,
Mass.: Blackwell
,
44-65.
B
ooij, Geert E. 1977.
D
utch Morpholog
y
.
Li
sse:
D
e
Ri
dde
r an
d
D
o
r
d
r
ec
ht
:
F
o
ri
s.
B
olozk
y
, Shmuel 1999.
M
easuring Productivity in Word Formation
.
L
e
i
de
n
:
Brill
.
C
arstairs-McCarthy, Andrew. 1992. Current Morpholog
y
.
London and New York: Routledge.
334
L
A
U
RIE BA
U
ER
C
homsky, Noam. 1957. Syntactic Structures
.
The Hague and Paris: Mouton.
C
homsky, Noam. 1964.
C
urrent Issues in Linguistic Theor
y
.
The Hague and Paris: Mouton.
C
homsky, Noam. 1970. “Remarks on nominalization.” In: R. A. Jacobs and P. S. Rosenbaum (eds),
R
eadin
g
s in En
g
lish Trans
f
ormational Grammar
.
Waltham
,
Mass.: Ginn
,
184-221.
C
orbin
,
Danielle 1987.
Morphologie dérivationelle et
structuration du lexique
t
.
2 vols. Tübin
g
en:
Nieme
y
er.
C
owie
,
Claire. 1999.
D
iachronic word-formation: a corpus-based
s
tudy of derived nominalizations in the
h
istory of English
.
U
npublished PhD thesis, University of Cambridge.
Dressler, Wolfgang U. 1997. On productivity and potentiality in inflectional morpholog
y
.
Clasne
t
Working Papers
7(
Montreal
)
.
Dressler, Wolfgang U. and Ladányi, Mária. 2000.
“
Productivity in word formation: a morphological
approach.”
A
cta Linguistica Hungarica 47
,
103-144.
Gie
g
erich, Heinz J. 1999. Lexical Strata in En
g
lish
.
Cambrid
g
e: Cambrid
g
e Universit
y
Press.
Harris, Zelli
g
. 1951. [Methods in] Structural Lin
g
uistics
.
Chica
g
o and London: Universit
y
of Chica
g
o
Pr
ess.
H
ay, Jennifer B. 2000. Causes and Consequences of Word Structur
e
. Un
p
ublished PhD dissertation,
Northwestern Universit
y
.
H
a
y
, Jennifer B. 2001. “Lexical frequenc
y
in morpholo
gy
: is ever
y
thin
g
relative?” Lin
g
uistics 39
,
1041-
1070.
J
es
p
ersen, Otto. 1942.
A
Modern English Grammar on Historical Principles. Part VI: Morpholog
y
.
London: George Allen and Unwi
n
, Copenhagen: Munksgaard.
Kastovsky, Dieter 1986. “The problem of productivity in word formation.” Linguistics
24,
5
8
5-
600
.
Kastovsky, Dieter. 1999. “Hans Marchand’s theory of wo
r
d-formation.” In: U. Carls and P. Lucko
(
eds.
)
,
Form, Function and Variation in English
,
Frankfurt-Main: Lang, 29-39.
Kiparsky, Paul. 1982. “Lexical mo
r
phology and phonology.” In: The Linguistic Society of Korea (ed.),
Lin
g
uistics in the Mornin
g
Calm
,
Seoul: Hanshin
,
3-91.
Knowles
,
Elisabeth 1997.
T
he Ox
f
ord Dictionar
y
o
f
New Words
.O
xford:
O
xford
Un
i
versit
y
Press.
Kruisin
g
a, Etsko. 1932. A Handbook o
f
Present-Da
y
En
g
lish
.
5th edition, Vol 3. Gronin
g
en: Noordhoff.
Kuiper, Koenraad. 1999. “Compounding by adjunc
t
ion and its em
p
irical conse
q
uences.” Languag
e
S
ciences 21
,
407-422.
L
ees
,
Robert B. 1963.
T
he Grammar o
f
En
g
lish Nominalizations
.
The Ha
g
ue: Mouton and Bloomin
g
ton:
I
ndiana University.
Marchand, Hans. 1969.
T
he Categories and Types of Present-Day English Word-Formation
.
2n
ded
n
.
M
u
ni
c
h
:
B
ec
k
.
Marle, Jaap van. 1985.
O
n the Paradigmatic Dimension of Morphological Creativit
y
.
D
o
r
d
r
ec
ht
:
F
o
ri
s.
Matthews
,
P.H. 1974.
M
orpholog
y
. Cambridge: Cambridge University Press.
Mayerthaler, Willi. 1981.
M
orphologische Natürlichkei
t
.
W
iesbaden: Athenaion.
Pla
g
, In
g
o. 1999.
M
orpholo
g
ical Productivit
y
. Structural constraints in En
g
lish derivation
.
B
e
rlin an
d
N
ew York: Mouton de Gru
y
ter.
Pla
g
, In
g
o; Dalton-Puffer, Christiane; and Baa
y
en, Harald. 1999. Morpholo
g
ical productivit
y
across
s
peech and writing.
E
nglish Language and Linguistics 3
,
209-228.
Q
uirk, Randol
p
h and Wrenn, C. L. 1957. An Old English Grammar
.
L
o
n
do
n
:
M
e
th
ue
n
.
R
affelsiefen
,
Renate. 1999
.
“Phonological constraints on English word-formation.” In: G. Booij and J.
v
an Marle
(
eds.
)
, Yearbook of Morphology 199
8
.
Dordrecht: Kluwer
,
225-287.
R
ainer, Franz. 1987. “Produktivitätsbegriffe in der Wo
r
tbildungstheorie.” In: W. Dietrich, H-M. Gauger and
H
. Geckeler
(
eds.
)
, Grammatik und Wortbildung romanischer Sprachen
,
Tübingen: Narr, 187-202.
Schultink
,
Henk. 1961. “Produktivit
e
i
t als morfolo
g
isch fenomeen.” F
o
r
um de
r L
ette
r
en
2: 110
-
12
5
.
Schultink, Henk. 1992. “Herkomst en groei van het begrip ‘produktiviteit’ als morfologisch fenomeen.”
I
n: M. Klein
(
ed.
)
, Nieuwe eskapades in de neerlandistie
k
,
Groningen: Wolters-Noordhoff, 187-201.
Smith, Logan P. 1912.
T
he English Language
.
Reset, 1944. London: Oxford University Press.
Štekauer
,
Pavol. 2000.
E
nglish Word-Formation. A History of Research (1960-1995
)
. Tübingen: Narr.
W
entworth, Harold. 1941. “The allegedly dead suffix -
DOM in modern En
g
lish.”
P
ublications o
f
the
M
odern Lan
g
ua
g
e Association 56
:
280-306
.
W
illiams
,
Edwin. 1981. “On the notions ‘lexi
c
ally related’ and ‘head of a word’.” Linguistic Inquir
y
12
,
245-274.
Zimmer, Karl E. 1964. Affixal Negation in Englis
h
and Other Languages. Supplement to Word 20.
335
Štekauer, P. and R. Lieber (eds.), Handbook of Word-Formation, 335—352
.
©
2005 Sprin
g
er. Printed in the Netherlands.
CONSTRAINTS ON PRODUCTIVITY
FRA
N
Z RAI
N
ER
1. INTR
O
D
UC
TI
O
N
A
pattern of word formation can never apply to
j
ust any word of a language, but
o
nly to some subset of words. This subset of possible bases is called the pattern’s
domain
.
In the case of unproductive patterns the domain can only be defined
extensionally, i.e. by enumerating one by one all the actually used bases, while the
domain of productive patterns must be defin
e
d
intensionally, i.e. by indicating one
o
r more features that any potential base
m
ust or should
p
ossess as well as additional
factors from outside the pattern itself that may be relevant. These features and
factors which serve to delimit a pattern’s domain are generally referred to in the
l
it
e
rat
u
r
e
a
s
const
r
aints
(
o
r
rest
r
ictions
,
when they are patter
n
-
specific). The exact
delimitation of the domain of particular patterns or, to put it the other way round, the
i
dentification of the constraints on those patterns is one of the most important and at
t
he same time difficult descriptive tasks students of word formation have to cope
with. The present chapter is intended to present a typology of constraints valid fo
r
n
atural languages in general, though illustrat
e
d mainly with examples from English.
2.
U
NI
V
ER
S
AL
CO
N
S
TRAINT
S
C
onstraints on patterns of word formation may be either universal or language-
s
pecific. We will start our discussion with universal constraints, of which again two
t
ypes can be distinguished, constraints supposedly located at UG (Universal
Grammar in Chomsky’s sense) and processing constraints.
2
.1 Constraints supposedl
y
located at U
G
From the end of the sixties onwards,
g
en
e
r
ative s
y
ntax was characterised b
y
an
endeavour to reduce the number of rules, placing at the same time powerful
u
ni
ve
r
s
al
co
n
s
traint
so
n th
e
m in
o
r
de
r t
o
a
vo
i
d
overgeneration
.
This research style
was taken over by generative morphologists when word formation once again
became a respectable branch in genera
t
ive grammar at the beginning of the
s
eventies. Over the next years, more than half a
do
z
e
n
suc
h
co
n
s
traint
swe
r
e
proposed, and
g
enerall
y
baptised with somewhat pompous names (for exhaustive
c
atalo
g
ues, see Scalise 1984: 137-165 and, in a more sceptic vein, Rainer 1993: 98-
116
)
: the Word Based H
y
pothesis at the heart of Aronoff (1976) and its corollar
y
,
th
e
No Phrase Constrain
t
– whose prohibition of phrasal constituents ironically was
t
already contradicted by the very name of this constraint! –, the Binary Branching
Condition
(
Aronoff 1976: 89 and 94; “one affix, one rule”
)
, a reformulation of the
336
F
RANZ RAINER
structuralist dogma of binarism, the Unitary Base Hypothesis
(
Aronoff 1976: 48
)
,
which banned any disjunction from the syntacticosemantic characterisation of the
d
omain of rules of word formation, later supplemented by Scalise (1984: 137) with
an analogous Unitary Output Hypothesis
,
the
A
djacency Condition (Siegel 1977)
an
d
th
e
Atom
C
ondition
(
Williams 1981
)
, which were designed to restrict the power
o
f rules to refer to constituents of the bases, to name
j
ust the most important
proposals. The most thorou
g
h anal
y
sis carried out in this research st
y
le b
y
a
g
enerative morpholo
g
ist was Botha’s (1988: 22-33) anal
y
sis of reduplication in
A
frikaans. Most of these constraints were flawed from the be
g
innin
g
b
y
an
i
nsufficient em
p
irical under
p
i
nning, and when the conditions-on-rules approach
went out of fashion in generative grammar, they met the quiet death they deserved,
which is not to say that some insights have not been integrated into later theories.
2
.2 Processin
g
constraints
W
hile
g
enerative
g
rammar up to the present da
y
has limited itself essentiall
y
to
t
he stud
y
of competence – as opposed to performance – and hence continues to
profess a thorou
g
hl
y
idealised and static conception of the lexicon, since the end of
t
he seventies we have been witnessing th
e
powerful development of psycholinguistic
s
t
ud
i
es
a
bou
t th
e
m
e
ntal l
e
x
icon (for a recent overview, see Baayen & Schreude
r
2003). While at the beginning morphologists and psycholinguists used to stand back
t
o back, one may observe more and more mutually fruitful cross-fertilisation ove
r
t
he last years, especially between psycholinguists and non-generative morphologists.
W
ith respect to our topic, some aspects of blockin
g
(2.2.1.), affix order (2.2.2.) and
c
hoice of the base(s) (2.2.3.) can onl
y
be fruitfull
y
tackled from a processin
g
perspective
.
2
.2.1 Blockin
g
Morpholo
g
ists speak of
blocking
when the unacceptability
g
of a morpholo
g
icall
y
c
om
p
lex word is not due to the failure to meet some re
q
uirement of the relevant
pattern of word formation but to the existence of either a synonymous word or a
synonymous pattern. In Rainer (1988) it was proposed to refer to the first of these
two c
a
ses
a
s
token blocking
and to the second one as
g
type blockin
g
.
T
oken blocking
I
n the case of token blocking, as we h
a
v
e already mention
e
d, the unacceptability
o
f one morpholo
g
icall
y
complex word is due to the existence of some s
y
non
y
mous
word in the lan
g
ua
g
e. The classic example is the unacceptabilit
y
o
f
stealer
as a
r
c
onsequence of the existence of
thief
. The requirement that a synonymous word
f
f
exist “in the language” is somewhat impreci
s
e, because what really matters is the
existence of an established synonym in the mental lexicon of the speaker or writer in
question, not in the language as a
s
ocial institution
(
de Saussure’s
l
angue
)
. In fact, it
c
an readily be observed that children, non-native speakers or other people with an
i
nsufficient command of the language produce complex words of the
stealer
type as
r
C
ONSTRAINTS ON PROD
U
CTIVITY
33
7
l
ong as they have not yet internalised the established synonym. Even adult native
s
peakers with a full command of their own language sometimes produce complex
wo
r
ds o
f th
e
stealer
type when they momentarily fail to retrieve the established
r
t
erm. This last observation makes it clear that blockin
g
has to do with processin
g
:
u
nder normal circumstances, lexical access priv
i
le
g
es established words, while word
formation is onl
y
resorted to when no established word is available or if the speake
r
o
r writer wants to avoid this established word for some stylistic purpose. The
blocking word need not be a simplex as in the case of
thief
, but may itself be a
f
f
r
egular complex word (c
f
. Malicka-Kleparsky 1985), provided that it is frequent
e
nough to be stored in the speaker’s or writer’s mental lexicon. Frequency is a
decisive conditioning factor for token blocking: the higher the frequency of the
blocking word, the stronger the blocking effect. In the case of low frequency words,
doublets are readil
y
tolerated.
W
hat is the status of the blocked word? It i
sc
l
e
ar that it m
us
t
be we
ll-f
o
rm
ed
in
principle accordin
g
to some productive patt
e
r
n of word formation of the lan
g
ua
g
e,
o
therwise it would not make sense to say
t
hat it ha
sbee
n
b
l
oc
k
ed.
N
eve
rth
e
l
ess
blocked words should not simply be put on the same footing as possible words, i.e.
c
omplex words that are not yet establish
e
d but readily acceptable
t
o native s
p
eakers.
R
aine
r
ian
,
for example, a relational ad
j
ective d
e
r
i
ved
fr
o
m th
e
nam
eo
f th
e
a
u
th
o
r
o
f
t
hese lines, would be such a possible but hitherto non-existent word. And since
possible words ma
y
serve as input for further derivation, nothin
g
could prevent us
from coinin
g
R
aine
r
ianism
should I ever become responsible of establishin
g
some
new political, philosophical o
r
o
ther trend. Now
,
we obse
r
v
e that blocked words
,
t
hough morphologically well-formed, behave quite differently in this respect fro
m
p
ossible words:
thief
, for example, does not only block synonymous
f
f
steale
r
,
but also
any further conceivable derivation thereof.
S
tealerless
,
for example, is just as odd as
stealer
itself, even though there is no established synonym
r
thiefless t
o
t
o
k
e
n-
b
l
oc
k
i
t, and well-formedness is guaranteed by the acceptability of
leade
r
less
,
teache
r
less
and similar formations. The same holds true for compounds:
piano thief
vs. *
f
piano
*
steale
r
,
etc. It mi
g
ht thus be useful
t
o replace the traditional dichotom
y
possibl
e
vs.
established
– or
actual
r
–
word
–
by the trichotomy
d
potential
vs.
l
possibl
e
vs.
established word
. According to this terminology, blocked words would be potential,
but not possible words, in our sense of words which are not only well-formed but
also readily acceptable under normal linguis
t
ic circumstances. This conce
p
tual
d
i
s
tin
c
ti
o
n all
ows us
t
o
f
o
rm
u
lat
e
t
h
e following general constraint on word
formation
(
in the seventies it would have been called Possible Base
C
onstrain
t
): The
t
bases of patterns of word formation must be possible words; or, put the other way
r
ound: merely potential – as opposed to possible – words are excluded as bases of
patterns of word formation.
Ty
pe blockin
g
I
n the case of type blocking, the unacceptability of a complex word is not due to
the blocking force exerted by an established synonymous
word
, but to the fact that a
s
ynonymous pa
tt
ern takes precedence. Type blocking
m
ay thus apply even when no
actual blocking word formed according to
t
he rival pattern exists. It will be
338
FRANZ RAINER
i
llustrated here with the rivalry bet
w
e
en the synonymous abstract suffixes -
heit
,
-
ität
an
d
-
ie
in German after adjectives ending in -
il
(
l
c
f
. Rainer 1988: 180-181
)
.
[
+final stress
]
-
heit
[
+final stress
]
&
[
+learnèd
]
& /-il/
-ität
[
+final stress
]
&
[
+learnèd
]
&
[
+ -
p
hil
]
-ie
S
chema 1
T
ype blocking with German quality nouns in
-heit, -ität
and
t
-
i
e
G
erman ad
j
ectives with final stress fall into the domain of the suffix -
heit
,
which
i
s full
y
productive and ma
y
even appl
y
to non-native words with final stress, as is
s
hown by
akut
‘acute’
t
ĺ
Akutheit
‘acuteness’,
t
galant
‘gallant’
t
ĺ
G
alanthei
t
‘gallantery’,
grottesk
‘grotesque’
k
ĺ
G
rotteskhei
t
‘grotesqueness’, etc. The domain
t
of
-
heit
,
however, is systematically
c
urtailed by the rival suffix -
ität
after learnèd
t
1
b
a
ses
in -
il
,
which all have final stress, as shown by
labil
‘unstable’
l
ĺ
L
abilität
‘
instability’,
ste
r
il
‘sterile’
l
ĺ
S
terilitä
t
‘sterility’, etc. It is important to note that
t
e
ven where no quality noun is attested in dictionaries, native speakers will clearly
p
re
f
e
r
-ität
to
t
-
heit
if the adjective has a learnèd flavour, such as
t
me
r
kantil
‘
mercantile’
,
monofil
‘made of one thread’, etc. Only
l
with some ad
j
ectives that have
become part of a more colloquial re
g
i
s
ter ma
y
one observe occasionall
y
that -
heit
is
t
al
so
t
o
l
e
rat
ed:
debil
‘stupid’
l
ĺ
D
ebilheit
‘stupidity’, beside established
t
Debilität
,
sku
rr
il
‘droll’
l
ĺ
S
kurrilhei
t
‘drollery’, beside established
t
S
kurrilitä
t
,
etc. We may
t
hus conclude that with learnèd ad
j
ectives with final stress -
ität
effectively type-
t
b
l
oc
k
s
-
heit
.
Now, it is interesting to observe that the domain o
f
-
ität
is again
t
c
urtailed by the even more specific domain of -
ie
with ad
j
ectives ending in the suffix
-phi
l
‘-phile’, which is of Greek origin. Adjectives in
l
-phil
should fall into the
l
do
main
o
f -
ität
since they are learnèd, end in
t
-
il
and have final stress, but
l
n
evertheless the
y
consistentl
y
take -
ie
:
b
ibliophi
l
‘book-loving’
l
ĺ
Bibliophilie
‘
bibliophily’,
anglophil
‘anglophile (adj.)’
l
ĺ
Anglophilie ‘anglophilia’, etc. The
s
ame would also be true of any neologism in -
p
hil
.
On the Internet, one may only
fin
dso
m
esc
att
e
r
ed
f
o
rmati
o
n
sw
ith -
ität
,
such as
Frankophilität
‘francophilia’ or
t
X
enophilitä
t
‘xenophilia’, which are probably be
t
s
t attri
bu
t
ed
t
o
a la
c
k
o
f
acquaintance with the learnèd vocabulary on the part of the coiners, but nevertheless
prove that
-phil
would fall in the domain of
l
-
ität
if this suffix were not type-blocked
t
b
y
-
ie
.
A
crucial question with respect to t
y
pe blockin
g
is whether it is predictable when
i
t applies. Van Marle (1986) ventured the h
y
pothesis that it should alwa
y
s appl
y
if
t
he rivalry obtains between a pattern with a relatively unbounded domain and a
pattern with a more restricted domain which is a subset of the larger domain.
U
nfortunately this generalisation is both too strong and too weak. It is too weak,
since
G
erman -
ität
shows that an affix need not be
t
a general case (a default) in orde
r
t
o be exposed to type blocking: even special cases may be further type-blocked by
o
ther more special cases. On the other hand, it is too stron
g
, since, as pointed out in
1
Note that [±learnèd] is a stylistic feature and not an etymological or morphological one like English
[
±Latinate
]
.
C
ONSTRAINTS ON PROD
U
CTIVITY
339
Rainer (2002: 119), one may also observe f
r
e
e variation between a more general and
a more specific pattern: in the case of Span
i
sh relational ad
j
ectives, for example, the
g
eneral pattern X
iano
and the more specific pattern
Xeano
ar
e
in fr
ee v
ariati
o
n
w
ith
forei
g
n names endin
g
in -
e
,
which is the specific domain of -
eano
.
So, for example,
f
r
om
Sartr
e
we ma
y
form both
sa
r
t
r
iano
an
d
sa
r
t
r
eano
,f
ro
m
d
e Saussur
e
,
saussu
r
iano
an
d
saussureano
,
etc. It is also eas
y
to ima
g
i
ne that one da
y
, due to an
ever more deficient mastery of learnèd vocabulary on the part of younger
g
enerations, -
ität
might come to be in free variation with
t
-
ie
aft
e
r
b
a
ses
in -
p
hil
.
Th
e
exact nature of the relationshi
p
between rival
p
atterns therefore seems to have to be
l
earned in every single case. This is another fundamental difference with respect to
t
oken blocking, where the positive correlation between the blocking force and the
frequency of the blocking word seems
t
o be language-independent. Strictly
s
peakin
g
, t
y
pe blockin
g
should therefore be classified under 3. in the present
c
hapter
.
2
.2.2 Complexity Based Ordering
A
n important processin
g
constraint affectin
g
the order of affixes, dubbed
Complexity Based Ordering
b
y
Pla
g
(2002), has been proposed recentl
y
b
y
Ha
y
g
(
2002). It sa
y
s that “an affix that can be easil
y
parsed out should not occur inside an
affix that cannot” (p. 528). Th
i
s processing constraint, acco
r
ding to Hay, explains
m
ost stacking restrictions attributed to level ordering in Lexical Phonology (cf. 3.1.
)
,
s
uch as the impossibility of attaching
-
it
y
to adjectives in -
less
:*
home-less-it
y
vs.
home
-
less
-
ness
,
etc. According to Hay’s constraint, this is simply due to the fact that
-
less
is more easily parsable than -it
y
an
d
h
e
n
ce s
h
ou
l
d
n
o
t
occu
r in
s
i
de
th
e
latt
er
affix. The de
g
ree of parsabilit
y
depends
,
amon
g
other factors, on the relative
frequenc
y
of base and derivative: the hi
g
h
e
r
the relative frequenc
y
of the base with
r
espect to that of the derivative, the hi
g
her the suffix’s parsabilit
y.
2
Accordin
g
to this
m
easure of parsability, -
ment
is more easily parsable in
t
improvement
than in
t
governmen
t
,
since
i
m
p
rove is more fre
q
uent than
i
m
p
rovemen
t
,
while
government
is
t
m
ore fre
q
uent than govern
.
This means that the degree of parsability of an affix may
v
ary from derivative to derivative. As Hay (2002: 549) shows convincingly, this
d
ifference in parsability explains why of our two words in -
ment
only the minimally
t
d
ecomposable derivative
government
may take the suffix
t
-
al
,
which is sensitive to
t
he presence of internal structure:
governmental
vs. *
l
i
mprovemental
.
This abilit
y
of
H
a
y
’s constraint to predict even intra-affixal variabilit
y
is certainl
y
a remarkable
a
c
hi
eve
m
e
nt
.
Now, is Hay’s constraint so powerful that we may dispense not only with level
o
rdering but also with affix-specific restrictions? The answer to this question,
according to Hay & Plag (in press), is n
e
g
ative: Hay’s processi
n
g constraint would
s
till all
ow
a
subs
tantial n
u
m
be
r
o
f
co
m
b
inations which are in fact unacceptable.
These can only be ruled out by the means of
affix-specific restri
f
c
tions. Hence, Hay
2
According to Hay & Baayen (2002), there is a positive correlation between parsability and productivity.
That means that more productive suffixes tend
t
o be located outside less productive ones.
340
FRANZ RAINER
and Plag conclude, “both selectional restrictions and processing constraints are
i
nstrumental in determining suffix ordering” (p. 28).
2.2.3 Productivit
y
,
f
requenc
y
and len
g
th o
f
bases
In general descriptive practice the domain of patterns of word formation is
described by indicating some salient features such as the word class of a potential
base, its semantic category, prosodic features, or whatever may be relevant in a
particular case (
cf
. 3.2.). As Krott, Schreuder, and Baayen (1999) have found out,
f
f
t
here are regularities beyond these combinat
o
rial specifications, totally unexpected
o
n the background of current models of word formation. They show that not every
word with the required specifications has
t
he same chance of becomin
g
a base fo
r
f
u
rth
e
r
de
ri
v
ati
o
n
o
r
o
f
en
t
erin
g
into a compound as a c
o
nstituent. The probabilit
y
in
fact is positivel
y
correlated with the frequenc
y
of a word, but ne
g
ativel
y
with its
l
ength and, in the case of a derivative, also with the productivity of the affix: more
fre
q
uent words are
p
referred to less fre
q
u
e
nt ones, short ones over long ones, and
words formed according to a less productive pattern over ones formed according to a
m
ore productive pattern. In view of the categories involved and the fact that the
c
onstraint cuts across all patterns of word formation it is quite plausible to assume
t
hat these constraints are a consequ
e
n
ce of general processing mechanisms.
3. LAN
GU
A
G
E-
S
PE
C
IFI
CCO
N
S
TRAINT
S
W
ith respect to lan
g
ua
g
e-specific constraints, one fundamental question is
whether the stackin
g
up of affixes is constrained onl
y
b
y
selectional
r
est
r
ictions
o
f
individual affixes – the null hypothesis – or
whether some more general principles
r
are involved. This hotly debated issue w
i
ll be taken up in 3.1. The second question
t
hat we will address is the nature (3.2) and typology (3.2.1-6) of affix-specific
res
tri
c
ti
o
n
s.
3.1 Level orderin
g
Massive borrowin
g
of complex words ma
y
lead to a situation where the word
formation patterns of a particular language are divided into two more or less tightly
s
eparated blocks, a foreign and a native one. Many scholars have argued that this is
e
xactly what has happened in English due to the massive influx of Romance and
L
atin words during the Middle English period. From a synchronic perspective, it has
l
ong been observed that affixes of Latin or Romance origin generally occur inside
affixes of Germanic origin: persuas-ive-ness
vs. *
h
ome-less-it
y
,
etc. The two types
o
f affixes also tend to differ in their phonolo
g
ical behaviour: affixes of Romance or
L
atin ori
g
in tri
gg
er more phonolo
g
ical chan
g
es in their bases (c
f
.
p
ersuad
e
/
p
ersuas-iv
e
above
)
and affect stress
(
c
f
.
átom
/
atómic
/
atom
í
cit
y
)
, while Germanic
affixes are stress-neutral
(
cf. hóme
/
hóme
-
less
/
hóme
-
less
-
ness
)
. These correlations
h
ave led Siegel (1979) to propose an organisation of English word formation in
blocks of rules in such a way that, roughly speaking, Latin and Romance affixes