?tekauer Pavol, Lieber Rochelle. Handbook of Word-Formation
Подождите немного. Документ загружается.
76
A
NDRE
WS
PEN
C
ER
counter-
p
roductiv
e
,
counte
r-
attack
the prefix
k
counte
r
-
combines with a noun
,
adjective and verb. On the other hand, the prefix
de-
in
de
-
ice
,
de-cipher
an
dsoo
n
r
seems to be responsible for turning a noun into a verb.
The original word syntax approach drew a distinction between individual words
and phrases. Such a model is able to acco
m
modate a principled distinction between
m
orphological and syntactic representations and the types of principles that apply to
t
hem, while retainin
g
the claim that morpholo
g
ical and s
y
ntactic principles can
o
verlap and even be shared (see Di Sciullo and Williams 1987 for detailed
d
iscussion). However, recent approaches to i
n
flection, inspired b
y
th
ewo
rk
o
f Bak
er
(1988) and Pollock (1989), have argued that all morphemes, whether lexical o
r
functional, head full
p
hrases, and that eith
e
r heads or
p
hrases can be incor
p
orated
i
nto word structures under certain circumstances. This marks essentially a return to
t
he position of Harris (1951). Although proponents of such a model sometimes
m
aintain the necessity for an independent morphology module it is unclear how such
a module would operate. I shall refer to such purely syntactic approaches to word
formation as ‘radicall
y
s
y
ntactic’ models.
The precise nature of the word s
y
ntax in radicall
y
s
y
ntactic models depends on
assumptions about s
y
ntactic structure. Those who follow the model of Ka
y
ne (1994)
ado
p
t an
X-bar model
of syntax, while those who follow the minimalism of
l
C
homsky (1995) adopt the
b
are
p
hrase structure a
pp
roach. I am not in a
p
osition to
say to what extent that approach is compatible with Kayne’s anti-symmetry model
.
The matter is further complicated by proposals advanced
r
e
cently within the
fram
ewo
rk
o
f
D
istributed Morpholog
y
(
Halle and Marantz 1993
)
, under which
l
exical items lack all lexical cate
g
or
y
specifications and acquire the features of noun,
v
erb or ad
j
ective
o
nl
y
in the s
y
ntax.
2
U
nfortunatel
y
, these proposals are as
y
et too pro
g
rammatic to permit serious
d
iscussion. I shall, however, briefly mention one influential set of proposals fo
r
c
ertain as
p
ects of word structure, nam
e
l
y, the model of argument structure as
l-
s
ynta
x
proposed by Hale and Keyser (1993, 2002). They consider denominal
c
onverted verbs in which the basic noun has the meaning either of a location (
shelve
the books
= ‘put the books on a shelf’) or a locatum (
saddle the ho
r
se
= ‘put a saddle
o
n a horse’). They propose that such verbs are derived from nouns by means of a
s
y
ntactic process of incorporation, akin to the kinds of incorporation structures we
see in noun incorporatin
g
lan
g
ua
g
es. Thus, (2) is said to be the structure underl
y
in
g
t
h
e
shelve
/
saddle
/
/
t
y
pe verbs:
2
I have been unable to locate a complete publish
e
d exposition of these claims. See Don (2004) for a
c
riti
c
al a
ssess
m
e
nt
o
f th
ec
laim
s
an
d
f
o
r r
e
f
e
r
e
n
ces to un
p
ublished handouts an
dwebs
it
e
a
dd
r
esses
w
hich
p
rovide further information.
W
ORD
-
FORMATION AND SYNTA
X
7
7
(
2
)
V´
V
VP
NP
V´
th
e boo
k
s
the horse V PP
P NP
s
h
e
lf
s
a
dd
l
e
The two verbs are derived by movement
processes which give the ‘surface’
t
s
tructures in
(
3
)
:
(
3
)
V´
V
V
P
V V NP V
´
P
V
the books
V
PP
th
e
h
o
r
se
N P P NP
s
h
e
lf
t
t
t
s
a
dd
l
e
However, it has been argued in great deta
i
l and very persuasively in Kiparsky
(
1997) that the crucial aspects of the grammar of location and locatum verbs (and
o
ther denominal verbs, including instrumental verbs such as
to hamme
r
)
are the
r
esult of world knowledge: saddles are (typically) put onto horses, not anywhere else
78
A
NDRE
WS
PEN
C
ER
and certainly not vice versa. Kiparsky points to a number of other difficulties with
t
he approach of Hale and Keyser (Hale and Keyser 2002, do not address these
o
bjections, and indeed Kiparsky’s article is not even in their bibliography). As
C
ulicover and Jackendoff
(
2005
)
point out there is very littl
e
empirical
j
ustification
for the structures Hale and Keyser propose. Presumably, the overriding motivation is
t
o account for as many as possible of the properties of words using exclusively
s
y
ntactic principles.
2
.3 Lexical integrity
The radically syntactic approaches to word formation bring with them the
i
mplicit assumption that there is nothing else of interest in word structure apart fro
m
syntax. However, the ma
j
ority view amongst morphologists is that word structure is
governed by principles that are in part distinct from syntactic principles (or, in the
c
ase of Anderson’s
(
1992
)
a-morphous theor
y
of word structure, completely
d
istinct). One important such principle is
l
exical inte
g
rit
y
.
Essentiall
y
, this notion is
c
haracterized ne
g
ativel
y
: pri
n
c
iples of s
y
ntax do not ha
ve
a
ccess
t
o
int
e
rnal
wo
r
d
structure. In particular, s
y
ntactic rules or principles cannot be responsible for the
co
n
s
tr
uc
ti
o
n
o
f infl
ec
t
ed o
r
de
ri
ved wo
r
ds.
Thi
s
m
e
an
s
that if la
be
ll
ed
tr
ee s
tr
uc
t
u
r
es
such as (2) are to be admitted then we may have to give interpretations to those
s
tr
uc
t
u
r
es w
hi
c
h mak
e
th
e
m rath
e
r
d
iff
e
r
ent from syntactic constructions proper.
Lexical integrity is a principle that can be adopted even on a word syntax approach,
j
ust as long as the syntactic representati
o
n
sw
hi
c
h a
ccou
nt f
o
r
wo
r
ds
tr
uc
t
u
r
es
ar
e
i
nsulated from the operation of processes in the s
y
ntax of phrases proper.
Several properties of word structure hav
e
been proposed as dia
g
nostics of lexical
i
nte
g
rit
y
, thou
g
h some of them are more reliable than others. For instance, it is
generally agreed that displacements which
r
ealize information structure
,
such as
t
opicalization/focussing, scrambling, wh-movement and so on cannot affect parts of
w
ords. Even productively formed noun-noun compounds in English, the most
phrase-like of English word types, seem to be immune from such processes, as is
exemplified in (4 - 7):
(
4) a. She would never
g
ive a morpholo
gy
lecture.
b. A morpholo
gy
lecture, she would never
g
ive
_____
.
c. She would never stud
y
morpholo
gy
.
d. Morphology, she would never study _____.
(
5) *Morphology, she would never give a _____ lecture.
(
6) a. What did she give?
b. A morphology lecture
(
7) a. What kind of lecture did she
g
ive?
b. A morphology lecture
c. *A morphology one
W
ORD
-
FORMATION AND SYNTA
X
79
d. *A morphology ___
e. *Morphology
However, there are anal
y
tic constructions, which look like lexical items (that is,
s
in
g
le ‘words’ in some sense) which clearl
y
consist of more than one s
y
ntactic word.
A
familiar example from En
g
lish is the verb-particle construction, as in
tu
r
nthe
l
ight ou
t
. In the most literal sense, analytic expressions such as
tu
r
n out
violate
t
lexical integrity, in that they consist of
two words which can be separated from each
f
o
ther in the syntax, and yet there is a clear sense in which non-com
p
ositional
expressions of this kind are single lexical
i
tems. These constructions have given rise
to a considerable literature, which I shall very briefly summarize here.
T
he range of particle constructions is
l
ucidly summarized by Jackendoff (2002a).
I
n the cases of interest here, the particl
e
s can appear either in pre-ob
j
ect position,
immediately after the verb (
put down the book
(
(
) or in post-object position (
k
k
put the
(
(
book down
). The post-ob
j
ect particle can co-occu
r
onl
y
with NP direct ob
j
ects, not
with PP com
p
lements: Jill grew up into a strong woman
,*
J
ill grew into a strong
woman u
p
.
In post-object position the particle can take a modifying element
(
s
p
ecifier
)
: eat those sandwiches right up
!
Verb-Particle-NP constructions are
distinguishable from Verb-PP constructi
o
ns. Thus, genuine particles cannot be
c
on
j
oined the way that prepositions can:
*
r
un up a bank overdraft and up a credi
t
card bill
as opposed to
l
r
el
y
on one’s
f
riends but not on one’s collea
g
ues
.
V
erb-particle combinations often have idios
y
ncratic meanin
g
s, so that the
y
have
t
o be treated at the ver
y
least as listed idioms. However, these items displa
y
other
p
ro
p
erties of lexical items in that certain derivational
p
rocesses seem to be
applicable to verb-particle constr
uctions. Jackendoff (2002a: 72) cites
r
r
no
minalizati
o
n
ssuc
h a
s
the rapid looking up of the information an
ddeve
r
b
al n
ou
n
s
(
l
ookup
,
see also Selkirk 1982 for discussion). In other languages we get more
systematic derivations. Acke
r
man and LeSourd
(
1997
)
illustrate this in some detail
w
ith the Hungarian preverb construction (similar to the English verb-particle
c
onstruction
)
.
Given this back
g
round we can see that there are some respects in which verb-
particle constructions resemble s
y
ntactic
c
onstructions and some respects in which
t
hey resemble morphology. Although a number of authors have attempted to provide
purely syntactic accounts within the minimalist program, these attempts can hardly
be said to be successful. In general, they are obliged to treat the basic construction as
c
onsisting of verb followed by a ‘small clau
s
e’ consisting of ob
j
ect NP and particle.
A
s Jackendoff points out (2002a: 90f), Engli
s
h does
ha
ve s
mall
c
la
use co
n
s
tr
uc
ti
o
n
s
but the particle constructions do not behave like them. The small clause approach is
l
ikewise criticised b
y
Ramchand and Svenoni
u
s (2001). The
y
ar
g
ue for a s
y
ntactic
approach which appeals to the Hale and Ke
y
ser (1993) notion of l-s
y
ntax,
i
ntroduced in section 2.2. As far as I can tell, this essentially reproduces the
c
onstructional analysis to be described below, but putting the semantic structure into
t
he syntax.
The most promising approach to the verb-particle construction is, perhaps, to
t
reat it as a ‘constructional idiom’
,
as argued by Jackendoff (2002a) and for Dutch
by Booi
j
(2002b). Jackendoff (2002a: 84) provides the following analysis for a
80
A
NDRE
WS
PEN
C
ER
specific type of expression, the ‘time-away’ construction as in
d
ance the night awa
y
(
Jackendoff 1997
)
:
(
8
)
a. Form [
VP
V NP
[
Pr
t
awa
y
]]
b. Meaning
‘
waste
[
T
im
e
NP
]
h
eedlessl
y
V
-
i
ng
’
More
g
enerall
y
, we can take (9) as a template for the verb-particle construction:
(
9
)[
VP
V NP
[
Pr
t
P]]
t
A
variety of meanings will then be associated with particular instantiations of P
(
and/or V), whether directional, aspectual or idiomatic. The fact that a single
c
onstructional type can be associated with a variety of meaning types is a reflection
o
f th
e
Separation Hypothesis (Beard 1995, Beard & Volpe, this volume). In effect,
t
his constructional analysis is the derivational equivalent of the analysis of auxiliary
+ participle constructions as periphrastic inflectional constructions
(
Ackerman and
W
ebelhuth 1998
)
.
C
onstructional approaches leave unclear the alternation between pre-ob
j
ect and
post-object particle positioning (or preverb position and other positions in the case
o
f Dutch, German, Hungarian and oth
e
r
languages). Each language has to be
c
onsidered on its own merits. As is clear from the work of Toivonen
(
2002, 2003
)
,
even closely related languages such as Swedish and Danish differ considerably on
t
hese constructions. Thus, Swedish has pre-ob
j
ect position particles but not post-
o
b
j
ect position particles. Danish, on the other hand has no pre-ob
j
ect particles.
Toivonen ar
g
ues that pre-ob
j
ect particles are non-pro
j
ectin
g
words, that is, elements
o
f a zero-level s
y
ntactic cate
g
or
y
that fail to pro
j
ect a phrasal constituent. In this
r
espect they are like syntactically represented clitics. They form a tight unit with the
preceding verb by virtue of
being adjoined at the X
f
0
level. This makes particles very
s
imilar to suffixes, except that they attach to
inflected words rather
than to stems. In
r
post-ob
j
ect position, however, the particles often do pro
j
ect phrases (in that they can
be preceded by modifiers such as
r
igh
t
,
completel
y
and so on). Thus, the analysis of
Toivonen requires that the lexical entry for one and the same element be given a
d
ual morphos
y
ntactic cate
g
orization.
H
ow can the existence of s
y
ntacticall
y
c
omplex lexical items such as verb-
particle constructions be reconciled with
t
he notion of lexical inte
g
rit
y
? Ackerman
and LeSourd (1997) discuss this question in detail. The starting point is the
o
bservation that it is the semantico-syntactic representation of an expression which
i
ndividuates it primarily as a single lexical
e
ntry. This means that a single lexical
entry should have a single semantic representation, and a single argument structure
(with the arguments being expressed as grammatical functions in a uniform way).
A
lternations in ar
g
ument structure or the expression of
g
rammatical functions (such
as the passive voice) are permitted to the extent that the
y
preserve meanin
g
and to
t
he extent that the
y
en
j
o
y
a fair de
g
ree of
g
eneralit
y
. In derivations which respect
(standard) lexical integrity we will typically see an affix creating a new lexeme
w
hich constitutes a single syntactic ter
m
i
nal, even though it is morphologically
W
ORD
-
FORMATION AND SYNTA
X
81
c
omplex. For example, the affixes of words such as
readable
o
r
thicken
s
h
ow
n
o
s
igns of syntactic independence.
However,
j
ust being part of a constructi
o
n type with idiosyncratic meaning is not
s
ufficient to
g
uarantee that we are dea
l
in
g
with a sin
g
le word in an
y
sense.
A
ckerman and LeSourd discuss the case o
f
complex predicates which realize an
f
ar
g
ument structure t
y
pe, such as the cau
s
ative construction. Standard examples
c
ome from Romance languages but up to a poi
n
t thi
sc
an
be
ill
us
trat
ed w
ith th
e
English
make
-c
a
us
ati
ve.
In
H
arriet made Tom eat the a
pp
le we can say that we have
an analytic construction make
N
P
V
P
,
in which the object NP of
make
f
u
n
c
ti
o
n
s
a
s
(controls) the sub
j
ect of the VP. However, there is no reasonable sense in which we
c
ould say that make
N
P
VP
is a morphological construct: it fails to behave like a
P
s
ingle lexical item for any of the purposes of morphology. The English
make
-
c
ausative must therefore be distin
g
uished, for instance, from those verb-particle
co
m
b
inati
o
n
sw
hi
c
h
c
an
se
r
ve
a
s
th
eb
a
se
f
o
r
de
ri
v
ati
o
n
.
Ackerman and LeSourd summarize this position as follows (1997: 99):
Lexical integrity does not hold of lexical items as such, but rather is a property of the
zero-level categories specified in lexical representations. Analytically expressed lexical
i
tems and syntactically derived expressions ar
e
alik
e
in
so
far a
sbo
th
co
n
s
i
s
t
o
f m
o
r
e
than one zero-level category.
T
hey modify the notion of lexical integrity by restricting it to a property of
X
°
t
erminals (s
y
ntactic atoms), not of lexemes. We can state this as follows:
(
10) Revised Lexical Integrity: syntactic rules cannot alter the lexical
m
eaning of words (including argument structure); syntactic rules have
no
a
ccess
t
o
th
e
int
e
rnal
s
tr
uc
t
u
r
eo
f
X
°
categories.
T
he typology of complex predicates proposed by Ackerman and LeSourd (1997:
100) is shown in Table 1, in slightly adapted form:
Standard affixation Verb-particle type Romance
c
ausative type
L
e
xi
c
al
i
nf
o
rmati
o
n
se
manti
cs
tr
uc
t
u
r
e
ar
g
ument structure
se
manti
cs
tr
uc
t
u
r
e
ar
g
ument structure
se
manti
cs
tr
uc
t
u
r
e
ar
g
ument
s
tr
uc
t
u
r
e
Morpholo
g
ic
al f
o
r
m
s
y
nthetic
m
orpholo
g
ical
o
b
j
ect:
[X Y
]
V
anal
y
tic
m
orpholo
g
ical ob
j
ect:
[
X
]
Prt
,[
Y
]
V
no
n-
m
orpholo
g
ical
o
b
j
ect:
[
X
]
V
,
[Y
]
V
Syntactic
Ex
p
ression
single syntactic
at
o
m
:
[XY
]
V
t
wo syntactic atoms:
[X
]
Pr
t
,
[Y
]
V
t
wo syntactic
at
o
m
s:
[
X
]
V
,
[Y
]
V
Ackerman and LeSourd’s typology
of complex predicates
y
8
2
A
NDRE
WS
PEN
C
ER
3. SYNTACTIC PHENOMENA INSIDE WORDS
In this section I briefly survey instances of violations of (morphological) lexical
i
ntegrity in English word formation. One frequent instantiation of this occurs with
l
oosely bound prefixes and neo-classical combining forms (‘prefixoids’) which can
be coordinated with each other
(
Strauss 1982
)
, as seen in
(
11
)
:
(11) a. pre- and even to some extent post-war (economies)
b.
p
ro- as o
pp
osed to anti-wa
r
c. hypo- but not hyper-glycaemic
By no means all prefixes or combining forms behave in this fashion, however:
(
12
)
a. *un- and re-tie
b. *in- or ex-port
c. *erithro- and leuco-c
y
tes
In
o
th
e
r in
s
tan
ces
I
co
nf
ess
t
o
v
acillating judgements, as with they need to be
poly- rather than mono-glots
.
It is difficult to think of
p
airs of suffix
e
s
that
cou
l
dbeco
ntra
s
t
ed w
ith
e
a
c
h
o
th
er
s
emantically in such a way as to permit natural coordination. However, even where
plausible putative examples can be concocted they turn out to be ungrammatical:
*
n
either joy-ful nor -less
.
Similarly we sometimes find affixes attaching to
c
oordinated stems. Jud
g
ements are sometimes variable but the followin
g
seem to be
all
ow
a
b
l
e:
(
13
)
a. write- or
p
rint-able
b.
m
ouse
-
o
r rat-lik
e
O
n the other hand, examples such as (14) seem to be completely excluded:
(14) a. *irrepair- and irreplace-able
b. *slow- and smooth-l
y
c. *milk- and cream-
y
Also, only limited types of coordinating conjunction are generally permitted,
preferably monosyllables. Thus, in contrast to the examples in (11), we would
c
ertainly not hear examples such as (15) even from a speaker who accepted (13):
(15) a. a component which is *rep
l
ace- but probably
n
ot repair-able
b.
a f
o
rm
w
hi
c
h
w
a
s
*m
ouse
- rath
e
r than rat-lik
e
Serious investi
g
ation of this question is
h
ampered b
y
two t
y
pes of indeterminac
y
.
First
,
it is difficult to decid
e
w
h
e
th
e
r
e
l
e
m
e
nt
ssuc
h a
s
p
re
-
or
h
ypo
-
are
p
refixes o
r
c
ompounding elements (see Strauss 1982 for discussion). Second, the syntax and
W
ORD
-
FORMATION AND SYNTA
X
83
se
manti
cs o
f
coo
r
d
inati
o
n i
s
far fr
o
m
cl
ear. Of relevance here is the general
phenomenon of Gruppeninflexion or ‘suspended affixation’ (Lewis 1967) found in
m
any languages. For instance in many langu
a
ges of the Altaic and Uralic families
i
nflectional suffixes realizin
g
number, case or possession can be distributed across
c
oordinated expressions, to
g
iv
e
co
n
s
tr
uc
ti
o
n
so
f th
e
f
o
r
m
pot- and pan-s
or
mothe
r-
and
f
ather-m
y
.
However, in
g
eneral this is onl
y
possible when sin
g
le words are
c
oordinated, not phrases, and only when the coordinated elements form a ‘natural
c
oordination’
(
in an intuitive sense
)(
see Wälchli 2003
)
.
A
nother type of deviation from strict lexical integrity is found when an affix
apparently attaches to a whole phrase, as in a why-does-it-have-to-be-me-ish
expression
.
I am not aware of any serious study of such formations, and their status
i
s unclear to me. A cursory internet search reveals large numbers of such coinings,
thou
g
h it also reveals that for some speakers
ish
has become a free morpheme with
r
ou
g
hl
y
the meanin
g
‘approximatel
y
’. A related kind of violation of lexical inte
g
rit
y
i
s found when entire phrases enter into compounds, in violation of Botha’s (1978)
N
o Phrase
C
onstrain
t
(a self-violating name), as in
t
a why-does-it-have-to-be-me
-
attitude
. In English (though not, it seems, in Dutch) such constructions sound
i
nfelicitous unless the embedded ex
p
ression
i
s very frequent and easily recognizable,
and this gives rise to the intuition that phrases cannot be compounded in this way
unless they are somehow seen as ‘fixed’ or
‘lexicalized’ (in some unclear sense).
r
4
. AR
GU
MENT
S
TR
UC
T
U
RE REALIZATI
O
N
4.1 Deverbal morphology
In Section 2.2 I mentioned Chomsky’s (1970) influential treatment of action
n
ominals as syntactic affixation. This has been the topic of enormous speculation,
both for English (see Malouf 2000b for one
r
ecent survey, and Roeper, this volume),
for other languages and comparable syntactic affixation analyses have been provided
for other types of derivation. I first discuss action nominals, then subject nominals
(‘a
g
ent nominals’) such as
driver
and finally
r
-
able
ad
j
ectives derived from transitive
v
erbs
,
such as
readable
.
This section is avowedl
y
descriptive, since theoretical
discussion of these phenomena is often empiricall
y
rather selective. In the followin
g
s
ection I briefly summarize the way such constructions can be treated in different
t
ypes of framework.
4.
1
.
1 A
ction nominals
A
c
ti
o
n n
o
minal
s
ar
ewo
r
ds de
ri
ved
fr
o
m
ve
r
bs w
hi
c
h ha
ve so
m
eo
f th
e
m
orphological and syntactic characteristics of nouns. Cross-linguistically there are
t
wo, in principle distinct, aspects to deverbal nominalization. On the one hand the
de
ri
ved wo
r
d
l
oses so
m
eo
f it
sve
r
b
al m
o
r
phosyntactic properties, while on the othe
r
h
and it gains certain nominal properties. Even in English these two aspects are
s
eparable. One feature of the base verb which is retained is closely related to the
84
ANDRE
WS
PEN
C
ER
v
erb’s semantics, namely the argument structur
e
or valency. A deverbal nominal
will regularly take complements which co
r
respond to the arguments of the base
v
erb. A hackneyed example is shown in (16):
(
16) a. The enemy destroyed the city
b. the destruction of the city by the enemy
c. the enem
y
’s destruction of the cit
y
d. the cit
y
’s destruction b
y
the enem
y
I
n (16b, c, d) we see the subject and object arguments of the base verb expressed
as prepositional phrases and as a preposed genitive phrase. In (16c) the genitive
phrase expresses the verb’s subject, while in (16d) the genitive phrases expresses the
o
b
j
ect. It should be mentioned at the outset that examples such as (16d) are
s
omewhat rare. Corresponding examples with other nominalizations are very poor:
(
17) a. *the desi
g
n’s improvement b
y
the en
g
ineers
b. *the book’s criticism b
y
the reviewe
r
c. *the poem’s translation b
y
the student
d. *the car’s repair by the mechanic
I
n these cases it would be perfectly po
s
sible to have a construction corresponding
to
(
16b
)
or
(
16c
)
: the improvement of the design by the engineers
,
the
r
eviewe
r’
s
criticism of the book
e
t
c.
k
In general the nominalization names either
the event or the fact of the event
r
h
appenin
g
. Thus, we can sa
y
(18) or (19):
(18) event reading
The destruction of the city by the enemy
occurred last Monday
The enemy’s destruction of the city
(19) propositional or factive reading
The destruction of the cit
y
b
y
the enem
y
appalled us
The enem
y
’s destruction of the cit
y
The examples given so far have involved latinate vocabulary with specialized
and idiosyncratic nominalization morphology:
d
estro
y
~
dest
r
uction
,
i
m
p
rove
~
i
mprovemen
t
,
t
r
anslate
~
t
r
anslation
,
repair
~
r
r
epair
.
However
,
there is anothe
r
type of nominalization which use
s
native affixational resources
,
the -in
g
n
ominalization. In principle all En
g
lish verbs have such a nominal, in the sense that
all En
g
lish verbs (without exception, sav
e
for the modal auxiliaries
)
have what
Huddlestone and Pullum (2002: 1173f) refer to as the ‘
g
erundive-participial form’ of
the verb
(
I shall call it the
‘
-ing
form’). Such forms can be used as straightforward
g
n
ominalizations, as in
(
20
)
:
W
ORD
-
FORMATION AND SYNTA
X
85
(
20) a. The Romans built the city
b. the building of the city by the Romans
c. the Roman’s building of the city
d. *the cit
y
’s buildin
g
b
y
the Romans
However, when a verb has its own idios
y
ncratic nominalization use of the -in
g
form is may well sound awkward or even be excluded:
(
21) a. ??the destroying of the city
b. ??the improving of the design
c. ??the translating of the book
d. ??the repairing of the ca
r
No
minalizati
o
n
sw
ith th
e
-ing
form specified by just the definite article often
g
h
ave a rather archaic feel to them
,
no
d
oubt because speakers have some awareness
that they were used much more often in earlier forms of the language
.
3
F
or
discussion of the historical development of the nominal and its morphosyntactic
properties see Westcoat (1994; also Malouf 2000b) and especially Hudson (2003)
an
d
th
e
r
e
f
e
r
e
n
ces c
it
ed
th
e
r
e.
A
frequently discussed property of
-ing
nominalizations is that they permit a
g
s
omewhat puzzlin
g
s
y
ntactic construction. Consider the examples in (22):
(22) a. John’s repaintin
g
of the fence (took
j
ust two hours)
b. (We were surprised at) John repainting the fence so quickly
c. (We were surprised at) John’s repainting the fence so quickly
In (22a) we see a nominal form which has nearly all the properties of a noun
except that it retains (semantically, at least) the arguments of the original verb. In
particular, the nominal
repainting
is preferentially modified
g
by ad
j
ectives and not by
a
dve
r
bs:
(23) a. John’s speed
y
repaintin
g
of the fence
b. *John’s speedily repainting of the fence
On the other hand, in
(
22b
)
we see a c
o
nstruction which is often regarded as
e
ssentially verbal (clausal): here the
-ing
form is modified by adverbs and not by
g
ad
j
ectives
:
(24) We were surprised at
a. John so speedil
y
repaintin
g
the fence
b. John repainting the fence so speedily
(
25
)
We were sur
p
rised at
3
They abound in the King James version of the Bible, e.g. Song of Solomon ii.12 ‘the time of the singing
o
f
b
ir
ds
i
sco
m
e
’
.