?tekauer Pavol, Lieber Rochelle. Handbook of Word-Formation
Подождите немного. Документ загружается.
270
WO
LF
G
AN
GU
.
D
RE
SS
LER
a
w
ar
e
n
ess
in th
e
f
o
rmati
o
n
o
f a
bb
r
ev
iati
o
n
s
o
f all types and the lack of a semantic
d
ifference between the input and the output of the abbreviatory operations of
c
lipping, acronym formation, etc. separate them (as being extragrammatical) fro
m
grammatical word formation.
Small children prefer higher degrees of iconicity even more than adults, not only
i
n the above-mentioned case of diminutives but also on the subparameter of
c
onstructional iconicit
y
, i.e. their first morpholo
g
ical rules are dia
g
rammatic
su
ffixati
o
n r
u
l
es.
2
.3 Indexicality preferences
A
nother
p
arameter derived from Peircea
n
semiotics is the
p
arameter of
i
ndexicalit
y
.
An index is a sign where the signans directly refers to the signatum. All
o
f morphology is indexical insofar a morphological marker refers to the base of the
ru
l
e
that intr
oduces
it
.
Thi
s
f
u
lfil
s
th
ese
mi
o
ti
cde
finiti
o
n
o
f an in
de
x
w
hi
c
h r
e
f
e
r
s
deicticall
y
to its nearb
y
ob
j
ect. Thus also on this parameter, affixation is more
n
atural than modification
,
because the i
nde
xi
c
al r
e
lati
o
n
be
t
wee
n an affix an
d
it
s
base is clearer, and much more so than conversion, which lacks an overt si
g
nans.
Moreover
,
in indexical relations
,
as al
r
eady claimed by the Greek Stoics and by
Peirce, adjacency is preferr
e
d
to distance. This
p
refere
n
ce
ha
sbee
n ra
d
i
c
aliz
ed
int
o
the status of a constraint, Siegel’s (1977) adjacency constraint/condition.
T
h
e
r
e
f
o
r
e
affixati
o
n
o
f a mark
e
r t
o
an i
m
mediately ad
j
acent base is preferred to
h
aving something inserted between an affix and its base, as is the case with
i
ntermediate interfixes, e.
g
. in Sp. pueblo ‘villa
g
e’
ĺ
d
imin
u
ti
ve
puebl-ito and
,
with
i
nterfixation
,
pueblo-ec-it
o
(
cf. Dressler & Merlini Barbaresi 1994
)
, where the
i
nt
e
rfix -
ec-
i
s
in
se
rt
ed be
t
wee
n th
eb
a
se
an
d
it
sd
imin
u
ti
ve su
ffix -
ito
. En
g
lish lacks
t
his unnatural morphological operation, as well as meaningless interfixes between
t
wo elements of a compound (as in German, Dutch and Slavic languages), unless
o
ne analyses the element -
o-
in
gas-o-meter a
s
an int
e
rfix
.
Another consequence of the ad
j
acency preference is the preference for fixed
m
orpheme order within a morphological word (due to stacking affixes to thei
r
r
espective bases), always maintained in English but not in very agglutinating
l
an
g
ua
g
es such as Turkish, Quechua and Mari.
Most of the compounds in all lan
g
ua
g
es have their head within the compound,
t
hus the
y
are endocentric compounds, such as in
blackboard
, which is a type of
board. Exocentric compounds have their head outsides or more precisely, the head
h
a
s
t
obe
inf
e
rr
ed.
Th
us
a
loud
-
mouth
is a person who, metaphorically, has a loud
m
outh
,
b
are-foo
t
an adjective derived from a noun-phrase,
p
ick-
p
ocke
t
i
s a
p
erson
who picks pockets, and a pass-por
t
was originally a document which allows to pass
a port. Endocentric compounds are by far preferred to exocentric compounds in the
l
anguages of the world, because they allow
muc
h
e
a
s
i
e
r a
ccess
t
o
the head, i.e. they
are more natural on the parameter of
i
ndexicalit
y
. Sometimes endocentric and
exocentric compounds are morpholo
g
icall
y
differentiated. For example, the plural
sab
r
e teeth
r
efers to teeth
(
endocentric
)
, the
p
lural
sab
r
e tooth
-
s
to animals having
s
abre teeth
(
exocentric
)
.
W
OR
D
-
F
ORMATION IN NAT
U
RAL MORPHOLOGY
271
2
.4 Pre
f
erence
f
or morphosemantic transparenc
y
F
rom the semiotic preference for transparenc
y
(Ko
j
1979), we can derive the two
parameters of morphosemantic and (2.5) morphotactic transparency. On the
p
arameter o
f
m
orphosemantic transparenc
y
,
full trans
p
are
n
c
y means fully
c
ompositional meaning, as is generally the case with inflectional meanings.
Motivation of a complex or derived word by its parts or by applying a word-
formation rule to its bases (or bases in case of compounds) is best served in case of
optimal transparency (on the pref
erence character cf. Rainer 2001).
ff
I
n word formation, morphosemantic transparenc
y
can never be complete,
because Fre
g
e’s principle
o
f semantic compositionalit
y
can hold onl
y
for s
y
ntax
where the meaning of a syntactic unit is ty
pically fully derivable from the meanings
o
f its syntactic constituents (except in idiomatic phrases, such as
to kick the bucket
).
t
This does not hold for word formation
,
i
n
s
ofar as all acce
p
ted words are stored and
t
hus lexicalised (Bauer 1983, Meyer 1992
)
,
whereas not yet accepted neologisms,
g
enerally, realise only one of the potential meanings of a compound or derivation.
Thus we must differentiate between transparent word formation meaning (G.
W
ortbildungsbedeutung, cf. Corbin’s 1987 notion of
sens const
r
uit
) and lexicalised
t
word meanin
g
. Hence word formation
r
ules can onl
y
predict word formation
m
eanin
g
but not the opacif
y
in
g
differences between word meanin
g
and word
formation meanin
g
.
The end
p
oint of opacification (= non-transparency) is fossilization, where
m
embers and their combination are hardly
v
isible, if at all. For exam
p
le, E.
lord
and
d
l
ad
y
are not recognizable as compounds any more, E.
n
ostril, gospel, handsel barely
for the first member
,
if at all. In contrast
,
E.
dandelion
m
ay be recognized as a
c
ompound, whose first member is difficult to identify, whereas its second is
semanticall
y
totall
y
opaque (diachronicall
y
a loan from Fr.
dent de lion
‚
lit. ‘tooth of
l
ion’
)
. Less lexicalization means more t
r
ansparenc
y
, more lexicalisation more
o
pacit
y
. More transparenc
y
implies more motivation of the compound via its
me
m
be
r
s.
D
ue to Frege’s principle of compositionalit
y
,
the meaning of a non-idiomatic
noun-
p
hrase, such as a high school
i
s fully compositional and thus transparent, the
c
ognate compound
high-school
is not. Still the compound high-school may be
l
c
lassified as a morphosemantically tra
n
s
parent compound, because the meaning of
t
he head is fully tra
n
s
parent and because the semantic
m
otivation by its first membe
r
(
non-head
)
is still evident: a hi
g
h-school
i
s hi
g
h in a metaphorical sense, i.e.
r
elativel
y
hi
g
h in contrast to elementar
y
and
g
rade school. In fact, a hi
g
h-school
is
a
s
pecific instance of what a ‘hi
g
h school’ ma
y
mean potentiall
y
. Thus, in a first
approximation, we may define an actual transparent compound as one whose
m
eaning is a subset of the set of potential meanings of the compound as constructed
g
rammatically via the combination of the meanings of the two parts. This is more
precise than Shaw’s (1979) criterion that the head must be an hyperonym of the
c
ompound. This is also relevant for psycholinguistic research on non-existing, but
potential compounds. The
y
are alwa
y
s transparent, but their use depends on the
possibilit
y
to instantiate via
(
e.
g
. metaphorical) inference
s
a pra
g
maticall
y
plausible
potential meanin
g
.
27
2
WO
LF
G
AN
GU
.
D
RE
SS
LER
B
ased on this approach and following Libben (1998) we can differentiate the
following four fundamental degrees of morphosemantic transparency (which are
c
learer and more systematic than Shaw’s 1979):
1) transparency of both members of the compound, e.g.
doo
r-
bell
,
2) transparency of the head member, opacity of the non-head member, e.g.
st
r
aw-
b
err
y
,
3) transparency of the non-head member, opacity of the head member, e.g. jail
-
bird
,
4
) opacity of both members of the compound, e.g.
h
um-bug
.
This scale of transparency presupposes t
h
at transparency of the head is more
i
mportant than of the non-head.
F
urther refinement is possible, if we diff
e
r
entiate between direct and indirect
,
i.e.
o
pacit
y
. Thus
ratchet tooth
h
as a transparent first member, but a metaphoricall
y
m
otivated second member (the leaves are compared with teeth), thus belonging to
t
he less trans
p
arent variant
o
f transparency degree 1).
The same procedure is possible with derivations. For example, the agent noun
undertak-er
belongs to an opaque variant of transparency degree 1), insofar as the
r
m
eaning of the lexical base undertake is specialised into a very specific undertaking.
I
n derivation, however, transparenc
y
of the head is much more important than in
c
ompoundin
g
. In fact, there appears to be an, at least implicit, unanimit
y
amon
g
m
orpholo
g
ists to speak of, for example, an a
g
ent noun onl
y
, if it is reall
y
an a
g
ent
n
oun, i.e. if the mor
p
hosemantic contribution of the suffix -
er
is transparent.
r
Otherwise, one speaks of instrument nouns (e.g.
print-er
of a PC, (
r
Geiger
)
count
-
e
r
)
o
r local nouns (e.g.
din
-
e
r
,
s
lee
p
-er
).
I
n principle, compounds are morphosemantically more transparent than
derivations, because they are more descriptive (cf. Seiler 1991), for example, the
c
ompound
d
ishwashing machine
t
ells the language user more about what it is, than
t
h
ede
ri
v
ati
o
n
dishwash
-
e
r
,
i.e. the compound serves better the function of semantic
m
otivation than the correspondin
g
derivation (cf. Crocco-Galèas and Dressler 1992).
Therefore, in the process of first lan
g
ua
ge
acquisition, several children have been
o
bserved to replace temporarily the l
e
s
s transparent (and ambiguous) agent and
i
n
s
tr
u
m
e
nt n
ou
n
open-er
with
r
o
p
en-man an
d
open-thing
respectively (Clark, Hecht,
g
and Mulford 1986
)
.
2
.5 Pre
f
erence
f
or morphotactic transparenc
y
On the parameter of
m
orphotactic transparenc
y
,
the most natural forms are those
where there is no opacif
y
in
g
obstruction to ease of perception. Purel
y
phonolo
g
ical
processes opacif
y
ver
y
little, e.
g
. res
y
llabification, as in
roast
ĺ
roaster
or the
r
a
pp
lication of com
p
ound-stress rules, as in
bláck bóa
r
d
ĺ
bláckboard
. More
m
orphotactic opacity occurs when morphonological rules intervene, such as in
conclude
ĺ
conclusion
,
even more so in cases o
f
allomorphic rules, as in
f
divide
ĺ
m
e
t
a
phor
a
a
i
c
m
o
t
i
va
t
i
o
n
(
cf. de Kno
p
1987
)
,
w
h
i
c
hwou
l
d
a
l
l
ow
t
od
i
v
i
de de
g
rees
1
)
–
3
)
above
i
n
t
o
s
ubdegrees
w
it
h
m
e
t
a
phor
a
i
ca
ll
y
m
o
t
i
v
a
t
ed vs. un
m
o
t
i
v
a
t
ed opac
i
t
y
.
W
OR
D
-
FORMATION IN NAT
U
RAL MORPHOLOGY
273
division
,
fiv
e
ĺ
fifth
,
b
r
oad
ĺ
b
r
eadth
.
Most opaque is suppletion, as i
n
th
r
ee
ĺ
third
(weak suppletion) and even more in
d
one
ĺ firs
t
,
two
ĺ
second
(strong
d
s
uppletion).
Another aspect of the preference for mo
r
photactic transparency is the preference
r
r
for continuous (rather than discontinuous) morphs. Therefore suffixation and
prefixation is preferred over infixation
(
discontinuous base
)
or circumfixation
(discontinuous affix). Thus diminutive/hypoc
o
ristic suffixes are generally preferred
to such infixes, as in S
p
.
C
esar-it
o
vs.
C
es-it-ar, hypocoristics of the name
C
esa
r(
cf.
Méndez Dosuna and Pensado 1990). English
h
as no infixation
,
other than within
e
xtragrammatical morphology of forming depreciatives, as in absolutely
ĺ
abso-
b
looming-lutel
y
. Circumfixes can occur only in languages which have a rich prefix
s
ystem (e.g. German and Dutch). Thus En
g
l
i
s
h ha
s
n
o
n
e
an
d
i
s
th
us
m
o
r
e
nat
u
ral in
t
his respect than those closel
y
related lan
g
ua
g
es.
B
ases of word-formation rules are, ceteris paribus, morphotacticall
y
and
m
orphosemanticall
y
most transparent, if the
y
are autonomous words in thei
r
uninflected form, already less so in their inflected citation form (Dressler 1988a, cf.
Rainer 1993: 98ff). This universal preference for word-based morphology applies to
c
ompounding even more than to inflection and derivation. Smaller bases than
autonomous words are more opaque and occur in English, in contrast to Latin,
Slavic and Semitic languages, very rarely, for example in the La
t
inate prefixed verbs
r
e-ceive, per-ceive, con-ceive, re-duce, ad-duce, con-duce
,
etc.
,
whose bases are
m
orphosemanticall
y
opaque an
y
wa
y
.
Larger and thus also universally less pref
erred bases are represented by inflected
f
f
wo
r
ds w
hi
c
h ar
e
n
o
t i
de
nti
c
al
w
ith
c
itati
o
n forms. Such bases are very rare and
r
estricted in English, e.g.
sp
ort-s-man, sale-s tax with a
p
luralised first base
(
cf.
J
ensen 1990). Pluralised second bases are wide-spread in Romance languages.
Still larger bases, namely phrases appear in synthetic compounds of the types
dish
-
wash
-
e
r an
d
lion-heart-ed
,
good-natur-ed
(
cf. Crocco Galéas 2003b
)
, in the
ty
p
y
e
,
t
h
t
t
ree
-
s
t
ar genera
l
,
t
h
t
t
r
ee-phase mo
t
o
r
,
three-color process
,
etc.
,
wher
e
t
h
e
m
odifier of the noun-phrase in the non-head position is often difficult to omit (e.
g
.
star
g
eneral belon
g
s to a different compound t
y
pe and desi
g
nates a ‘
g
eneral who is a
s
tar’). For other phrasal-compound patterns cf. Lieber (1992: 92ff) and Booi
j
(1992:
4
5ff). Whole sentences as first members of a binary compound occur only rarely,
suc
h a
s
in
d
o-it-yourself movement
,
w
hi
c
h i
s
l
ess
nat
u
ral than eye movement.
C
omplex sentences occur only in consci
o
usly formed occasionalisms, such as
an oh-
what
-
a
-
wicked
-
wo
r
ld
-
thi
s-is-and-how-I-wish-I-could-do-something-to-make-it-
i
i
b
etter-and-nobler-expression
(
J. K. Jerome: Three Men in a Boat
)
.
T
here is an interaction between the parameters of iconicity and transparency.
Of
t
f
f
e
n we f
i
f
f
nd
in
t
he s
a
m
e
t
y
pe of
y
word
f
for
m
a
t
i
o
n a
t
e
n
denc
y
t
o
w
ar
ds
i
c
on
i
c
it
y
between morphosemantic and morphotactic transparenc
y
/opacit
y
. For example
c
omparable nominal compounds tend to be morphotactically opaque when they are
m
orphosemantically opaque. For example, the first base is morphosemantically
m
ore trans
p
arent in
mother-land
and
d
main-land
th
a
n in
d
Dixie-land
. In analogical
parallelism in morphotactics, the phonological shape of the second base has more
r
arely a reduced (ant thus morphotactci
a
l
ly more opaque) vowel in the first and
s
econd than in the last compound.
274
WO
LF
G
AN
GU
.
D
RE
SS
LER
2
.6 Preference for biuniqueness
Another semiotically-based parameter has
b
iuniqueness a
s
it
s
m
os
t nat
u
ral
o
ption. Biuniqueness, which holds if one and the same form alwa
y
s has the same
m
eanin
g
(and vice-versa), is more natural than uniqueness and especiall
y
ambi
g
uit
y
.
Agglutinating languages have much mor
phological biuniqueness,
r
r
En
g
lish ver
y
little
(
cf. Bauer 1983, Plag 2003). For example, the suffixes
-
able
,
-less and many
p
refixes are biuni
q
ue,
a
l
though biuniqueness is relative even here because -
able
ha
s
the un
p
roductive allomor
p
h -
ible
and the meaning of -
less
overla
p
s with that of the
prefix
un-
.
Examples of uniqueness (one-to-many relations) is the formation of ordinal
nu
m
be
r
sv
ia th
esu
ffix -
th
,
but -
th
also forms nouns from adjectives, as in
wid
-
th
,
b
r
ead
-
th
,
l
en
g
-th
.
Or the suffixes -it
y
an
d
-
ness
h
ave no other function than formin
g
n
ouns from ad
j
ectives, but the
y
are competitors.
Examples of ambi
g
uit
y
(man
y
-to
-
man
y
relations) are the suffixes -
er
(agent and
r
i
nstrument and local nouns, comparative vs. agent nouns in -
ist
,
-
ato
r
)
, -
al
and
l
-
ant
(
forming nouns and adjectives), -
en
(forming adjectives and verbs, also a plural
m
arker), always in competition with
o
ther suffixes
(
cf. also Plag 2004).
B
iuniqueness can be obtained more easily in terminology (Felber & Budin 1989:
122f). However even here terminologists do not strive for obtaining general
biuniqueness but only terms wh
i
ch are biunique within each scientific domain (in
the horizontal stratification of lan
g
ua
g
es for specialized purposes), particularl
y
fo
r
the most abstract level of theoretical ar
g
umentation (in the vertical stratification of
LSP
)
. Thus it is un
p
roblematic that a term such as
a
r
ticulation
h
as a very different
m
eaning in phonetics and in medicine, as long as it has always the same meaning in
p
honetics and if this
p
honetic conce
p
t is not
e
xpressed by another concept as well.
S
ciences where different terms are indiscri
m
inately used for the same concept (as
o
ften in pedagogy) are likely to get a low rating for such usage. Since, however,
different schools of a discipline often tend to use the same terms differently, the
quest for biuniqueness must be limited to the same text world, i.e. biuniqueness of a
t
e
rm
s
h
ou
l
d
h
o
l
d
at l
e
a
s
t
w
ithin th
es
am
e
text
(
cf. Dressler 1994: 956f
)
. In addition
we find a tendenc
y
to enhance biuniqueness of the morpholo
g
icall
y
decomposed
parts of words, such as compound members, e.g. in attributing the sense of an
applicative descriptive science to the combining for
m
-graph
y
,
as in
l
exico-graph
y
,
termino-graphy
vs.
l
exico-log
y
,
termino-log
y
.
2
.7 Fi
g
ure-
g
round pre
f
erences
A
nother universal preference is the tendenc
y
towards
figure-ground
sharpenin
g
d
(
Scherer 1984), i.e. for contrastin
g
s
y
nta
g
maticall
y
between a more important, more
d
y
namic and clearer fore
g
round and a less important, more static and fuzzier
background. In word formation the head represents the figure, the non-head the
g
round. Therefore, as we have seen in section 2.4, morphosemantic transparency of
the head is more im
p
ortant than that of the non-head. This is also another reason
why endocentric compounds are preferred to exocentric ones (cf. section 2.3).
W
ORD
-
FORMATION IN NAT
U
RAL MORPHOLOGY
275
T
he figure-ground distinction between head and non-head is established by
m
orphosemantic and morphotactic subordination of a non-head under the head (cf.
s
ection 2.2). This is by far the pref
e
r
red situation in suffixation and compounding.
T
he onl
y
partial exception in suffixation is diminutive formation (and au
g
mentative
and pe
j
orative formation in other lan
g
ua
g
es), as in
d
o
gg
-ie
,
where the suffix does
n
ot determine the class of the base (a t
y
pi
c
al head propert
y
), unless in rare instances
suc
h a
s
b
r
own
-
ie
, where the suffix transforms an adjective into a noun.
T
he
p
reference for subordination of a non-head to a head holds even in
exocentric compounds. Not only is the non-head morphosemantically subordinated
t
o the absent head. But subordination holds, secondarily, even among the actually
present members of exocentric compounds:
loud
is subo
r
d
inat
ed
t
o
mouth
i
n
loud-
mouth
,
whereas in pick-pocke
t
we find again non-uniformity of (secondary) head-
t
h
ood: althou
g
h the primar
y
, semantic head which desi
g
nates the person who picks
pockets, is not expressed, pocke
t
i
s, secondaril
y
, the morphotactic head which
d
etermines inflection, whereas s
y
ntacticall
y
pocke
t
is subo
r
d
inat
ed
t
o
pic
k
.
I
n compounding, subordinate compound
s
are universally preferred ove
r
c
oordinate com
p
ounds which have two or more mor
p
hosemantic heads
(
cf.
W
underlich 1986: 241), i.e. without the clear figure-ground distinction of
s
ubordinate compounds. Thus in
s
peaker-hearer both members are of equal status,
although the plural ending attaches only to the right member, another example of
n
on-uniform headhood. Coordinate (or coordinative) compounds ma
y
a
g
ain be
endocentric
,
such as speaker-hearer
o
r the ad
j
ective
bitte
r-
sweet
,
also called
appositional compounds, or the
y
ma
y
be exocentric, such as morpholo
gy
-s
y
ntax
i
nterface, where the two coordinated com
p
ound members have their semantic head
ou
t
s
i
des:
it
co
in
c
i
des w
ith
i
nterface
,
the syntactic head
of the whole noun-phrase.
d
More subtle
p
ro
p
erties of coordinat
e
compounds may differ considerably fro
m
l
anguage to language (cf. Olsen 2001): The
l
in
e
ar
o
r
de
r
o
f m
e
m
be
r
s
in
coo
r
d
inat
e
c
ompounds is not grammatically determined (since all members are equipollent), but
pragmatically (e.g. the most important first) or stylistically, e.g. prosodically (e.g.
t
he lon
g
est last). The first reason explains the order of
s
peaker-hearer
(
becaus
e
l
in
g
uists tend to think more of the speaker than of the hearer, cf. the ter
m
native
s
peaker), the second explains wh
y
the order of the s
y
non
ym
s
peaker-listener
is eve
n
m
ore difficult to reverse
(?
l
istener-s
p
eaker
)
than in the case of
sp
eaker-hearer
.
On the syntagmatic level, this lexical-pragmatical grading of importance in
sp
eaker-heare
r
is antagonistic to the inverse morphological order of non-heads
r
followed by heads in English and the ma
j
ority of languages. The latter sequence
r
epresents a universal preference for heads to be on the right side of non-heads,
c
alled the righthand head rule by William
s
(1981: 248). A minority of languages has
l
ar
g
e classes of lefthand-headed compounds as well (cf. Zwanenbur
g
1992a, b,
Scalise 1992: 179ff, Rainer 1993: 57
)
. This
r
ecalls the suffixin
g
preference, whereb
y
suffixes are
p
referred to
p
refixes. Most
e
xplanations of the suffixing preference start
from the assum
p
tion that it is better for a word to start with the lexical basis
(
cf. Hall
1992), which would hold for both right-headed and left-headed compounds. Thus if
o
ne compares compounds, then the more valid generalisation seems to be that it is
better for complex words to end with the head. This would also explain the tendency
for prefixes not to be heads (cf. Hall 1992). An explanation for the right-hand head
276
WO
LF
G
AN
GU
.
DRE
SS
LER
preference may lie in the psycholinguistic
r
ecency effect which makes the end of a
word more salient, which is especially im
portant in early language acquisition (cf.
m
m
W
ijnen et al. 2001).
2
.8 Pre
f
erence
f
or binarit
y
Semiotic and thus also
g
rammati
c
al relations are preferentiall
y
b
inar
y
,
as alread
y
o
bserved b
y
Peirce (1965: II.277), a preference which is ma
y
be based on the binar
y
n
ature of neurolo
g
ical information transmittance), and this both in paradi
g
matics and
s
yntagmatics. As has been ascertained al
r
eady within structural linguistics, even
apparently ternary relations usually have to be split up into two binary relations. Fo
r
e
xample, morphology has to be subdivided into inflectional morphology and word
formation, the latter in its turn into derivation and compounding.
In syntagmatic relations, the preferred patterning consists in concatenating one
e
lement to one base, for example by prefixing or suffixing to only one base, whereas
i
n circumfixin
g
(less natural) two element
s
ar
e
affix
ed
t
oo
n
eb
a
se.
Th
es
am
e
h
o
l
ds
for compoundin
g
, both in subordinate and coordinate compoundin
g
. This preference
h
olds true for coordinated com
p
ounds such as
q
ueen-mother
or
p
rince-consor
t
,
but
(due to entirely extralinguistic reasons) not for flags, e.g.
red
-
white
-r
ed
f
o
r th
e
Austrian and Peruvian flag. Subordinate three-member compounds usually have to
be grouped into two binary relations, e.g. [[[three][star]][general]], where the phrase
[[
three
][
star
]]
forms an intermediate unit.
2
.9 Optimal shape o
f
units
Th
e
o
p
timal extension of a
g
rammatical morpheme, be it an autonomous one
s
uch as pronouns or of a bound morpheme (prototypically an affix), is one syllable.
The optimal extension of a simplex lexical word is one foot, which means in
c
ombination with the prefere
n
c
e for binarity (2.8), a bisyllabic foot. The binarity
preference also entails that one word has only one derivational or inflectional affix.
If these consist of no more than one syllable, then the resulting word shape lies
optimally still within the limit of a trisylla
bic foot. English derivational affixes are
a
i
ndeed protot
y
picall
y
monos
y
llabic, such as the suffixes: -
cade
,
-
cy
,
-
dom
,
-fold
,
-
f
ul
,
-hood
,
-
let
,
-
like
,
-l
y
,
-
ment
,
-
ness
,
-shi
p
,
-
scape
,
-
some
,
-ward
,
-wise
,
and
without the s
y
llable onset: -
age
,
-
al
,
-a/enc
e
,
-a/en
t
,
-
ate
,
-ed
,
-
ee
,
-
eer
,
-
en
,
-
er
,
-
ese
,
-es
q
u
e
,
-
ess
,
-
ette
,
-
(
i
)
an
,
-ie/
y
,
-
ic
,
-ing
,
-
ish
,
-
ism
,
-
ist
,
-
ive
,
-i
z
e
,
-
ous
,
-
u
r
e
.
S
malle
r
shapes, such as the coda, are represented by -
c
,
-
th
.S
uffixes which extend over more
t
han one syllable are: -
able
,
-ar
y
,
-
ation
,
-enc
y
,
-
er
y
,
-
ical
,
-if
y
,
-
itis
,
-
it
y
,
-te
r
ia
.
2
.10 Alternative naturalness parameters
The naturalness parameters of 2.2
–
2.8 differ partiall
y
from Ma
y
erthaler’s
(
1981), which are essentiall
y
maintained b
y
Wurzel (1984 and later).
W
ORD
-
FORMATION IN NAT
U
RAL MORPHOLOGY
27
7
For example, instead of the parallel pair of morphotactic and morphosemantic
t
ransparency, Wurzel has one single parameter of ‘morpho-semantic transparency’
with several subparameters. One of them is called ‘semantic transparency
(compositionalit
y
)’ and corresponds to
t
he parameter of morphosemantic
t
ransparenc
y
(2.4), whereas the subparameter of ‘morpholo
g
ical (formal)
t
ransparenc
y
’ corresponds to morphotactic transparenc
y
(2.5).
Mayerthaler’s parameter of uniform sy
mbolisation corresponds, on the one hand,
y
y
t
o the
p
arameter of biuni
q
ueness
(
2.6
)
. On
t
he other hand, it is partially covered by
t
he subparameter of the optimal base of a morphological rule (2.5).
W
hen comparing such alternative solutions, the following criteria are relevant:
which solution is more consistent, both internally and with the theory of naturalness?
W
hich has a more solid extralinguistic base, including psycholinguistic evidence?
W
hich one is terminolo
g
icall
y
more transparent? Which one is empiricall
y
more
s
atisfactor
y
?
Furthermore
,
which solution allows a better functional explanation? Thus
m
orphological naturalness is connected to the main function of morphology which
c
onsists in word-internal mor
p
hosemantic and mor
p
hotactic motivation.
W
hich solution allows better intercom
p
onential com
p
arisons, i.e. which conce
p
ts
h
ave wider application in Natural Phonology, Natural Te
x
tlinguistics (cf. Dressle
r
1985b, 1990)? Thus preferences for iconicit
y
,
indexicality, transparency (incl.
c
ontinuit
y
), (bi)uniqueness, and binarit
y
have been shown to hold in othe
r
c
omponents as well.
Finall
y
, there is the criterion of covera
g
e: Ma
y
erthaler’s (1981) three parameters
o
f constructional iconicity, transparency and uniformity have much less explanatory
c
overage than the seven preferences of 2.2 – 2.8.
2.11 Predictions and conflicts
The above-sketched deductive subtheory of
universal morphological naturalness
f
preferences allows several predictions (cf. Ma
y
erthaler 1981, Dressler 1985c):
U
niversal preferences should have a si
g
nificant effect on cross-lin
g
uistic
distribution. And in fact, more natural options on one parameter occur more
frequently than less natural ones on the same parameter (cf. Dressler 1982). If such a
frequency prediction does not hold cross-linguistically, then it is either falsified or it
m
ust be shown that diminution of statistical significance is due to naturalness
c
onflicts, i.e. to conflicts between universal preference parameters within the
m
orphological component itself or between preferences of different components, fo
r
example between morphology and phonology, a perspective put forward already by
t
he Neo
g
rammarians of the 19th centur
y
in
t
heir conceptualization of the diachronic
c
onflict between sound law (G. Laut
g
esetz) and analo
gy
(cf. Wurzel 1988b).
Morphosemantic transparenc
y
(1.4) holds much more in inflectional morpholo
gy
t
han in word formation. Dressler (e.g. in Dressler et al. 1987) has tied this to a
biuniqueness conflict within word formation, i.e. to a conflict between the tendency
t
owards biunique form-meaning relations of morphemes vs. those of whole words.
F
or example, in flut-is
t
,
morpheme biuniqueness should hold separately for the
278
WO
LF
G
AN
GU
.
D
RE
SS
LER
m
eaning and form of a) flute and b) agent noun formation with -
ist
,
whereas whole-
word biuni
q
ueness holds for the whole com
p
lex word flutis
t
.S
ince words as units
h
ave a semiotic priority over morphemes, lexical (whole-word) biuniqueness wins
o
ut and accounts for idiosyncratic lexicalisations in word formation. This is,
predictably, not the case in inflectional
m
orphology, because inflected words are not
l
exical entities in their own right.
The mea
g
er distribution of biuniqueness
v
s. uniqueness and ambi
g
uit
y
in
g
eneral
i
s due to another conflict, viz. with econom
y
, i.e. parsimon
y
of stora
g
e (cf.
Ronneber
g
er-Sibold 1980), a
g
eneral co
g
nitive factor.
More s
p
ecific
p
redictions hold for domai
n
s of external evidence. Thus
,
in
diachronic change, more natural options shoul
dbe
m
o
r
es
ta
b
l
e
than l
ess
nat
u
ral
o
n
es
o
n one and the same naturalness
p
aramet
e
r
, and the direction of morphological
c
hange should be preferably towards more
m
orphological naturalness. In addition to
n
aturalness conflicts
(
such as the above-mentioned
)
, there is a need for a
s
pecification of types of morphological change (cf. Wurzel 1994, Dressler 2002b).
Or in the domain of lan
g
ua
g
e acquisiti
o
n, children should identif
y
and acquire
m
orpholo
g
ical rules the earlier the more natural these rules are accordin
g
to
universal parameters (cf. above 2.2, 2.4).
B
ut clearl
y
naturalness conflicts are of a
different nature in children than in adults
(
cf. Dressler & Kar
p
f 1995
)
.
3. TYPOLOGICAL ADE
Q
UACY
Inspired by basic insights of Skali
þ
k
a (1979) on ideal language types which are
approached by natural languages to a greater
or smaller degree, and which consist of
r
properties which favour one another, languag
e
types have been reinterpreted within
N
M as
(
alternative
)
sets of consistent
r
esponses to naturalness con
f
licts
.
Since not
t
he most natural o
p
tions on all
p
arameter
s
can be combined within one lan
g
ua
g
e,
naturalness on certain parameters must be, so to sa
y
, sacrificed for
g
reate
r
naturalness on others
(
cf. Dressler 1985c, 1988a
)
.
Thus the inflecting-fusional type (as approached by Old English) combines
o
ptimal values on the universal parameters of binarity (2.8), of optimal shape (2.9),
figure-and-ground sharpening (
2
.
7) and indexicality (2.3)
w
ith less than optimal
v
alues in constructional iconicity (
2
.
2), morphosemantic and morphotactic
t
ransparency (2.4, 2.5) and biuniqueness (2.6). Rather the opposite holds for the
a
gg
lutinatin
g
t
y
pe, as well represented b
y
Turkish. The isolatin
g
t
y
pe has ideall
y
no
g
rammatical morpholo
gy
, especiall
y
no inflection, but ma
y
have much
e
xtra
g
rammatical morpholo
gy
, and has a preference for monos
y
llabic words.
Modern English morphology presents a rather peculiar typological mix. Its
m
orphological components or submodules approach different ideal types: English
i
nflectional morphology is rather isolating, which is also the true for its greater
preference for monosyllabic base words than is the case for other Germanic
l
anguages (except Afrikaans). Another impact of the isolating type lies in its very
productive conversions. The Germanic stra
t
um of its derivational morphology is
r
ather a
gg
lutinatin
g
(in its
g
reat iconicit
y
and transparenc
y
, i
n
c
ludin
g
the preference
for words as bases
)
. In contrast, the Lati
n
ate stratum of its derivational morpholo
gy
W
ORD
-
FORMATION IN NAT
U
RAL MORPHOLOGY
279
approaches the inflecting-fusional type (especially in its allowance of less iconicity
and transparency). Finally the ric
h
ness of English compounding recalls the
polysynthetic-incorporating type, including the allowance of noun incorporation as
i
n the s
y
nthetic verbal compounds of the t
y
pe
to
dish
-
wash
.
Thus universall
y
rather unnatural options
m
a
y
be t
y
polo
g
icall
y
adequate if the
y
fit the properties of the respective lan
g
ua
g
e t
y
pe (in the framework of the subtheor
y
o
f typological adequacy). In contrast to universal, ‘paradigmatic’ and autonomous
n
aturalness (section 2.), typological naturalness thus is relational, in that it relates
d
iff
e
r
e
nt
u
ni
ve
r
s
al nat
u
ral
n
ess
p
arameters and various
l
anguage components to each
o
ther. System-dependent naturalness (4) is then relational within the language-
s
pecific means and operations of a single component, e.g. morphology.
4
.
S
Y
S
TEM-DEPENDENT NAT
U
RALNE
SS
4.1 System-adequacy
L
anguage-specific, system-dependent naturalness, as conceived by Wurzel
(1984) for systems of inflectional morphology, represents what is normal or system-
adequate within the morphology of a language, although it may contradict some
u
niversal morphological preference (as delineated in 2). Among competing system-
defining structural properties the most dominant is the most adequate one.
For both the word stock and derivational morphology of English it is system-
s
pecific that the
y
are not simpl
y
stratifi
e
d
accordin
g
to a feature [+/– forei
g
n], as
m
ost other lan
g
ua
g
es (cf. Rainer 1993: 129f), but accordin
g
to the feature [+/
–
L
atinate] (cf. Plag 2004: 84). Thus suffixation with -it
y
i
s
r
es
tri
c
t
ed
t
o
Latinat
e
bases
,
whereas suffixation with -
ness
i
s
n
o
n-Latinat
e.
N
o
n-Latinat
ede
ri
v
ati
o
n i
s
dominant
,
thus -
ness
may also be attached to Latinate bases
,
as in o
p
a
q
ue-ness a
s
a
v
ariant
o
f
o
pac-it
y
.
The most important aspect of dominance is productivity. Morphological
productivity, on the level of the potential system of grammar, can still be defined in
H
enk Schultink’s wa
y
, as translated b
y
v
a
n
Marle (1985: 45) = ‘the possibilit
y
fo
r
l
an
g
ua
g
e users to coin, unintentionall
y
, a number of formations which are in
principle uncountable’ (cf. also Bauer 2001).
As with many other concepts of naturalness theory, productivity is also gradual.
This gradualness corresponds – similar to inflectional morphology – to the following
h
ierarchy of linguistic criteria (cf. Dressler & Ladányi 2000):
a
)
Wurzel’s
(
1984
)
second
a
r
y productivity in the inte
g
ration of unfitting foreign
words (where unfitting properties must be fitted, i.e. adapted to the indigenous
s
ystem),
b) Wurzel’s (1984) primary productivity in the integration of foreign words with
already fitting properties,
c
) derivation from indigenous abbreviations and other extragrammatically formed
bases
,