?tekauer Pavol, Lieber Rochelle. Handbook of Word-Formation
Подождите немного. Документ загружается.
THE LEXICALIST APPROACH TO WOR
D
-
FORMATION
16
7
A
n insightful method to treat the inherently compositional nature of word-
formation processes was proposed by All
e
n
(1978) applying the ‘IS A’ Condition to
English compounds like the following
:
1
9
(
30) [[high]
A
[school
]
N
]
N
T
he ‘IS A’ Condition allows the identification of the compound’s head in
s
emantic and categorial terms:
•
IS
high school A type o
f
high
o
r IS it A type of
school
?
•
G
iven that high is an adjective and
school
is a noun: IS
l
high school
A
N
adjective or IS it A noun?
A
high school ‘IS A’ (type of)
school
and it ‘IS A’ noun
,
hence
school
is the
l
h
ead of the compound
h
igh school
.
Looking back at the compounding rule in (11c),
w
e can generalize that in an English compound [X+Y]Z, Z ‘IS A’ Y, both from a
c
ate
g
orial and a semantic point of view.
A
lthou
g
h ‘IS A’ ma
y
correctl
y
tell us which
o
f the constituents is
p
redominant in
a complex word, it says nothing about the mechanisms that bring about this
asymmetry. Lieber (1980) presents a system
of rewriting rules th
m
at generate binary-
branching tree structures whose terminal nodes are filled by stems and affixes,
d
epending on their subcategorization frames. In this system the ‘IS A’ Condition is
explained on strictly formal terms: the essential property of a morphological head is
t
hat all its features (whether semantic or categorial) are copied to the upper node of
t
he structure. Lieber reformulates the ‘IS A’ Condition as the F
eatu
r
e
-P
e
r
colation
Convention
.
We illustrate this in sim
p
lified form with the wor
d
h
a
pp
iness
:
(
31) a. The features of a stem are passed to the first dominating non-
A
happ
y
A
n
ess
N
b. The features of an affix are passed to the first dominating node
w
hi
c
h
b
ran
c
h
es
N
A
happ
y
A
n
ess
N
1
9
This notion had been previously explored, though in different terms, for example by Marchand (1969).
168
S
ER
G
I
OSC
ALI
S
E
&
EMILIAN
OGU
E
V
ARA
B
oth the ‘I
S
A’
C
ondition and the Feature-Percolation
C
onvention were
d
esigned to account in formal terms for one of the most crucial notions in
m
orpholo
gy
: the notion of
head
. Aronoff (1976) extended the use of ‘head’ to
i
n
c
l
ude bo
th
wo
r
ds
an
d bou
n
d
f
o
r
ms
(
affixes
)
. Williams
(
1981
)
further elaborated it
,
proposin
g
the Ri
g
ht-Hand Head Rule
(
RHR) for En
g
lish: all morpholo
g
icall
y
c
om
p
lex words are headed, a
n
d the head in a com
p
lex st
r
ucture is the rightmost
element. The validity of his rule can be tested with prefixed, suffixed and compound
words
(
32a
)
and can be ex
p
ress
e
d with a general schema (32b):
(
32
)
a. [re+ [write]
V
]
V
[[happy]
A
ness
]
N
[[
black
]
A
[
board
]
N
]
N
b.
Y
XY
r
e
w
rit
e
happy ness
b
la
c
k
bo
ar
d
I
n all these cases, the ri
g
htmost
e
l
e
m
e
nt h
e
a
ds
th
eco
n
s
tr
uc
ti
o
n
:
rew
r
ite
i
s
a
ve
r
b
bec
a
use
w
r
ite
is a verb
,
ha
pp
iness i
s
a n
ou
n
bec
a
use
-
ness
forms nouns and, finally,
blackboard
is a noun because
d
board
is a noun.
d
A
logical generalization stemming from the RHR is that prefixes are neve
r
h
eads, while suffixes always are. This generalization, although correct for a great
n
umber of cases, is too strong and needs to be weakened: for instance, some prefixes
d
o indeed seem to be heads
(
cf.
rich
ĺ
en
+
rich
)
while evaluative suffixes and
i
nflectional affixes in
g
eneral do not qualif
y
as heads (cf.
book
+
let
,
want+ed
). As
d
d
for compounds, subsequent research has shown that the RHR is not a universal
principle but that it depends on the
ty
polo
g
ical affiliation of the lan
g
ua
g
es
c
onsidered: for instance, compounds in the Romance languages are systematically
l
eft-headed
(
cf. It.
uo
m
o
r
ana
‘lit. man-frog, frogman’)
.
20
The very concept of ‘head’ has been
s
trongly criticized. An extreme position,
d
efended by Bauer (1990), sugg
e
sts that the possibility that
the notion of head has
t
n
o place in morphology should b
e
al
so
tak
e
n int
o
a
ccou
nt
.
2
1
Nevertheless
,
the notion of head is
s
t
ill a fundamental one in morphology for it
g
rants important insi
g
hts into the functionin
g
of man
y
re
g
ular phenomena: the head
i
n a complex word is usuall
y
the locus of inflection (cf.
h
i
g
h school
ĺ
hi
g
h schools
20
Note that while derived words are always
e
n
docentric and by and large right-headed, compounds
e
xhibit a variet
y
of possibilities: (
a
)
head in the ‘canonic’ location
(e
.g
. ri
g
ht for Germanic lan
g
ua
g
es,
l
eft for Romance languages), (b) no heads (cf. the ‘exocentric compounds’, also called
bahuv
r
ihi
,
such
a
s
pale fac
e
)
,
(
c
)
two heads
(
dvandva
com
p
ounds, cf. S
p
.
poeta-pintor
‘poet-painter’), (d) head in the
r
‘
wrong’ place (cf. It.
te
rr
emoto
‘earthquake’ a right-headed word due its Latin diachronic origin).
21
Even the position taken by Bauer has been criticized. Cf. Štekauer
(
2001
)
.
THE LEXICALIST APPROACH TO WORD
-
FORMATION
169
an
d
It
.
capostazion
e
‘
s
tati
o
n ma
s
t
e
r’ ĺ capistazione
)
, it is the constituent fro
m
w
hi
c
h r
e
l
ev
ant inf
o
rmati
o
n percolates to the uppe
r
node, its position in compounds
i
s strongly related to the word-order type of the language (SVO languages tend to
h
ave left-headed compounds, while SOV lan
g
ua
g
es tend to have ri
g
ht-headed
o
n
es
22
).
Morpholo
g
ical theor
y
has often benefited from pro
g
ress in s
y
ntactic theor
y
. Fo
r
exam
p
le, Selkirk
(
1982
)
23
extended the hierarchical
pr
o
jections of the so-called
X
-bar schema to apply in morphology
:
24
the maximal morphological projection is
ide
nti
c
al t
o
th
e
X
0
level projection in syntax (i.e. words)while
,
below
X
0
,
two purely
m
orphological projections exist: Root (
X
-
1
)
and Af. The internal structure of
happiness an
d
monst
r
ous
would, therefore, be represented as in (33):
(
33
)
Word
(
N
0
)
Word
(
A
0
)
Word
(A
0
)
Affix Root
(A
-1
)
Root
(
A
-
1
)
-ness Root
(
N
-
1
)
Affix
happ
y
monstr- -ous
The similarity between morphology and syntax, however, is not complete. Fo
r
i
nstance, affixes do not have X-bar level: they are a peculiarity of W-syntax (the
syntax of words, morphology), having no parallel in S-syntax (syntax proper).
I
n addition, while in S-syntax it is common that non-maximal pro
j
ections
dominate other maximal pro
j
ections,
2
5
Selkirk argues that the same is not true of
m
orphology: she proposes a universal principle that, only in the domain of
W
-s
y
ntax, no constituent can dominate a constituent of hi
g
her X-bar level. This
m
eans that Words ma
y
contain Roots, but not vice versa; at the same time, both
W
ords and Roots ma
y
freel
y
contain Affixes, for the
y
do not have X-bar level.
22
En
g
lish, bein
g
an SVO lan
g
ua
g
e with ri
g
ht-headed
c
ompounds, is a counterexample to this statement.
H
owever, this can be explained in diachronic t
e
r
ms: English right-headed compounding is a remnant
o
f an earlier dominant SOV word order. It is wid
e
l
y accepted that there was a
s
hift in
wo
r
do
r
der
between Old and Middle English: Old English was SOV (like Proto-Indo-European) while Middle
English was mainly SVO (cf. Kemenade 1987, Lightfoot 1991).
23
Selkirk, while remainin
g
committed to the Lexicalist
hy
pothesis, presents a n
e
w
point of view: rathe
r
than highlighting the differences between morphology
and syntax, she explores the degree to which
y
t
hey resemble each other. The extent of this si
m
i
larity in Selkirk’s framew
o
rk
co
n
ce
rn
s
th
e
f
o
rmal
apparatus of both components. W-sy
ntax and S-syntax make use of
the same context-free rewriting
f
r
ules and of the lexical categories N, A and V.
24
Cf. section 7 below for other issues concernin
g
the application of X-bar theor
y
to morpholo
g
ical
p
henomena.
2
5
Cf.
[
the
[[
train
]
N
0
[[[
to
]
P
0
]
P
1
[[[
London
]
N
0
]
N
1
]
NP
]
P
P
]
N
1
]
NP
, where, for example, the projection N
1
h
eaded b
y
t
r
ain
contains a maximal pro
j
ection of the preposition
to
(
PP
)
.
170
S
ER
G
I
OSC
ALI
S
E
&
EMILIAN
OGU
E
V
AR
A
5
.1 Stron
g
and Weak Lexicalism
Stemming from Chomsky’s (1970) proposal that semantically irregula
r
de
ri
v
ati
o
n
s
h
ou
l
d
n
o
t
be
accounted for by the syntax (the beginning of Lexicalism),
t
here developed two opposed theoretical positions.
Th
e
S
tron
g
Lexicalist H
y
pothesis
t
akes Chomsk
y
’s su
gg
estion to its extreme
c
onsequences, excludin
g
all morpholo
g
ical phenomena from the s
y
ntax: the
processes of word formation and the rules of inflection are applied pres
y
ntacticall
y
,
i
n the Lexicon. This position was origi
n
ally proposed in Halle’s (1973) seminal
paper on generative morphology and it has bee
n
widely assumed as part of the most
i
nfluential theories of syntax (e.g. Lexical Functional
G
rammar
,
G
eneralized Phrase
S
tructure
G
rammar
,
H
ead-Driven Phrase
S
tructure
G
rammar
,
and to some extent
–
at least regarding inflectional morphology and checking theory – also by the
Minimalist Program).
The Stron
g
Lexicalist H
y
pothesis is usuall
y
supplemented b
y
the assumption that
s
y
ntactic rules cannot modif
y
, move or delete parts of words: this is the so-called
Principle o
f
Lexical Inte
g
rit
y
, which has been endorsed b
y
a
g
reat number of
m
orphologists (cf. Lapointe 1980, Di Sciullo & Williams 1987, among others) and
c
onstitutes one of the key ideas of
S
trong Lexicalism
.
This
p
rinci
p
le can be defined
as
(
34
)
, in its most radical formulation:
(
34) Generalized Lexicalist H
y
pothesis (Lapointe 1980: 8)
No s
y
ntactic rule can refer to elements of morpholo
g
ical structure
The Strong Lexicalist Hypothesis demands a sharp division between syntax and
m
orphology and, as such, it cannot account for a variety of phenomena that require
some degree of interaction of these two components of the grammar.
Th
ed
i
s
tin
c
ti
o
n
be
t
wee
n
de
ri
v
ati
o
nal m
or
phology (realized in the Lexicon) and
r
r
i
nflectional morpholo
gy
(accomplished b
y
the s
y
ntax) achieves a
g
reater de
g
ree of
d
escriptive accurac
y
. This position is known as the Weak Lexicalist H
y
pothesis an
d
(
in general) it has had more success among morphologists than among syntacticians.
A
ronoff (1976) acknowledged a Weak Lexicalist framework, as did many others
t
hereafter. The stand
p
oint received a
de
tail
ed
a
ccou
nt f
o
r th
e
fir
s
t tim
e
in
A
nderson’s classic pape
r
Where’s Morphology
?
(
1982
)
.
A
ll in all, over a ten-year period, the Lexicalist approach in morphology
d
eveloped a very rich body of hypotheses and principles, of which only a few have
been discussed in the precedin
g
sections. W
e
ha
ve
n
o
t m
e
nti
o
n
ed
m
an
y
others such
a
s
th
e
Ad
j
acenc
y
Condition (Sie
g
el 1978),
26
which develo
p
ed into the A
tom
Condition of Williams (1981), inheritance of Ar
g
ument Structure (Sproat 1985,
B
ooij 1988), etc.
Lexicalism has been a
pp
lied to an eve
r
growing number of languages and topics
r
suc
h a
s
th
e
nat
u
r
eo
f th
eso
-
c
all
ed
evaluative suffixes and clitics (Zwicky 1985),
26
Which inspired a great number of proposals, such as Pesetsky’s (1979)
B
racket Erasure Convention
,
W
illiam’s
(
1981
)
Atom
C
ondition
,
La
p
ointe’s
(
1981
)
for his Lexical Integrity Hypothesis
,
Kiparsky’s
(
1982
)
B
racketing Erasure Principl
e
,
Botha’s
(
1981
)
M
orphological Island Constrain
t
,
among others.
THE LEXICALIST APPROACH TO WORD
-
FORMATION
171
allomorphy (Carstairs McCarthy 1987), bracketing paradoxes (Spencer 1988), the
o
rganization of so-called non-linear mo
r
phologies (McCarthy 1982), the interface
between different components of the grammar (for example, between morphology
and s
y
ntax, see to name but one Zwick
y
1986; or between morpholo
gy
and
phonolo
gy
, cf. Booi
j
& Rubach 1987).
C
ertainl
y
, such an empirical ferment over a short period of time also brou
g
ht into
t
he
p
icture data that could not be acco
m
modated in the original theory, and
t
herefore
,
different branches of lexicalis
m
develo
p
ed. To be sure, Lexicalism was
n
ot the only theory of morphology: at the same time other theories developed such
as Natural Morphology (cf. Dressler et al
.
1987), to mention but one, giving rise to a
r
ich theoretical debate, which is still today very intense (cf. for example, Fradin &
Kerleroux 2003
)
.
6
. M
O
RE
O
N THE N
O
TI
O
N
O
F LEXI
CO
N
R
esearch in theoretical lin
g
uistics has remarkabl
y
enriched the notion ‘Lexicon’:
b
y
movin
g
awa
y
from the earl
yg
enerative viewpoint (cf. section 3), much more
elaborated hypotheses of lexical representations have been proposed. These
d
evelopments often involve the inclusion of information depicting the Pr
edicate
Argument Structure
(
PAS) and/or some kind of complex semantic description o
f
l
exical items (usually termed Lexical Conceptual Structure
,
LCS). Typical minimal
entries in such a Lexicon could be represented as follows (cf. Lieber 1992: 22):
(
35
)
a.
enter
[V ___ ]
r
[
¥'
n
t
'
(
r
)
]
LCS:
[
Event
GO
t
(
[
Thin
g
]
,
[
Path
TO ]
(
[
Pla
ce
IN
(
[
Thin
g
]
)
]
))
]
PAS: x <
y
>
b.
cat
[N ___ ]
t
[
¥
k
æ
t
]
LCS:
[
Thin
g
]
F
or instance, in
(
35a
)
enter
reads as a verb pronounced [
r
ε
ntr
]
whose LCS
desc
ri
bes
an a
c
ti
o
n in
vo
l
v
i
n
g two entities, one of which goes into the other;
ente
r’
s
PAS says that it has two arguments, x (external) and y (internal and, in this case,
n
on-obli
g
ator
y
).
Given such a conception of lexical representations and of the operations that can
be a
pp
lied to them, the domain of mor
p
hol
o
gy (traditionally und
e
r
stood as the study
o
f the internal structure of words) has been extended to include also the study of the
external valency of words (the effects of morphology on PAS
27
)
.
2
7
Whether there is an inde
p
endent PAS slot in lexical entries is open to question; some linguists assume
t
hat PAS is derived from LCS, as a sort of “summar
y
”
of its contents as far as it visible to s
y
ntactic
processes.
17
2
S
ER
G
I
OSC
ALI
S
E
&
EMILIAN
OGU
E
V
ARA
W
hile it is generally agreed that words are represented in the Lexicon (i.e. they
h
ave a full lexical entry of the kind seen above), whether affixes are lexically
r
e
p
resented is much more controversial. These issues demand a
p
rinci
p
led formal
d
istinction between ‘word’ and ‘affix’, and this is not a simple distinction to draw.
There are two basic viewpoints:
(
36
)
a. words and affixes are lexical items
(
both have full lexical entries
)
b. words and affixes are different (only words have entries in the
Lexicon
)
A
mong others, Halle (1973), Lieber (1980) and Selkirk (1982) ascribe to the
hy
pothesis (36a), which has been supported b
y
numerous ar
g
uments, for example:
•
W
ords and affixes often exhibit the same relations among them (synonymy,
antinomy, hyponymy, polysemy, cf. Lehrer 1996)
•
Sometimes, a unit clearl
y
identifiable as an affix on formal
g
rounds seems to
carry the kind of meanings expressed by
roots in other languages (cf. Mithun
y
1996
)
•
Bo
th
wo
r
ds
an
d
affix
es
ha
ve
l
e
xi
c
al
c
ategory and subcategorization frames (cf.
Lieber 1980, 1992, Williams 1981, Selkirk 1982
)
.
•
B
oth words and affixes take part in X-bar structures (Selkirk 1982), suffixes, in
particular, seem to share the basic prop
e
r
ties of s
y
ntactic heads in complement-
h
ead structures
(
Di Sciullo 1995
)
.
Hypothesis (36b) was first proposed in a Lexicalist framework by Aronoff
(
1976) and has become the hallmark of word-based theories of morphology: only
‘words’ are represented in the Lexicon, ‘affixes’ are assimilated to rules and operate
i
n a different submodule of the
g
rammar. The ar
g
uments for this position are also
n
umerous; its main a
pp
eal lies in the re
p
resentation of non-concatenative
phenomena such as umlaut, allomorph
y
, suppletion, all of which cannot be easil
y
explained by hypothesis (36a) as combinations of words and affixes. Furthermore
,
as
D. Corbin (1987) pointed out, if affixes and words have the same representation,
t
here would be no possible distinction to be drawn between compounding and
d
erivation: crucially, though, in some
l
anguages compounds and derived words are
s
ystematically different (e.g. compounds in the Romance languages are left-headed,
while derived words are ri
g
ht-headed, cf. It.
uo
m
o
r
ana
‘lit. man-fro
g
, fro
g
man’ vs.
ba
r
ista
‘bar man’
)
. If words and affixes are not differentiated, interestin
g
g
eneralizations such as this one will be lost.
I
t ha
so
ft
e
n
bee
n
obse
r
ved
that th
e
L
e
xi
c
o
n may contain a wide range of different
entities, produced not only by morphological processes but also by syntactic
o
perations. Di Sciullo & Williams (1987: 14
)
propose a “hierarchy of listedness” fo
r
t
he contents of the Lexicon, which they consider to be “like a prison – it contains
o
nly the lawless, and the only thing that its inmates have in common is lawlessness”:
(
37
)
– All the mor
p
hemes are listed.
– ‘M
os
t’
o
f th
ewo
r
ds
ar
e
li
s
t
ed.
THE LEXICALIST APPROACH TO WOR
D
-
F
ORMATION
173
– Many of the compounds are listed.
– Some of the phrases are listed.
– F
ou
r
o
r fi
ve o
f th
ese
nt
e
n
ces
ar
e
li
s
t
ed.
This does not mean that we cannot distin
g
uish between morpholo
g
ical and
sy
ntactic ob
j
ects; Di Sciullo & Williams emphasize the Lexicon’s role as stora
g
e fo
r
whatever linguistic object needs to be memo
r
ized by speakers. The elements of the
L
exicon for Di
S
ciullo
&W
illiams are all called
listemes
(whether they are words,
affixes or
p
hrases
)
.
7
. LEXI
C
ALI
S
M T
O
DAY
The Lexicalist Hypothesis in its strong version is rather difficult to maintain with
r
espect to a series of counterexamples that have been highlighted by empirical
r
esearch. These countexamples lead some lin
g
uists to conceive the morpholo
g
ical
c
omponent as havin
g
some
obli
g
ator
y
interaction with the s
y
ntax. The extent of this
i
nteraction, as conceived in various studies, ran
g
es from the so-called Weak
L
exicalist Hypothesis (which assigns inflectional morphology to the syntactic
c
omponent), to the opening of some systematic areas where morphology and syntax
“
talk” to each other, to the complet
e
a
ccou
nt
o
f all m
or
phological phenomena by
rr
m
eans of syntactic operations (which amounts to the exact opposite of lexicalism).
B
elow we will discuss some problems of the lexicalist approach and evaluate to
w
hat
e
xt
e
nt it
c
an
s
till
be co
n
s
i
der
ed a
p
roductive theoretical framework.
Stron
g
Lexicalist models are essentiall
y
linear
:
28
the morpholo
g
ical component
(Lexicon + Word Formation Rules) provides simplex and complex words, feedin
g
t
he structures created by syntax. The
o
nly point of contact between the two
c
omponents in these models is lexical insertion, the mechanism by which the
terminal nodes in a syntactic tree are ‘f
illed’ with words. The following is a
f
f
s
implified schema of this relation:
(Phonology,
Lexical Semantic
Insertion Interpretation)
````
Figure
3
In a picture such as this one, communication between morphology and syntax is
k
ept to a bare minimum. The Lexicon feeds the initial point of s
y
ntactic derivations,
2
8
There are also other types of models. Cf. for example the so-called ‘parallel morphology’ (Borer 1991,
Sadock 1991
)
.
Morphology
(
Lexicon +
W
FRs
)
S
y
nta
x
1
74
S
ER
G
I
OSC
ALI
S
E
&
EMILIAN
OGU
E
V
AR
A
l
eaving phonological and
s
emantic interpretation to take place after the syntactic
c
om
p
utation is done.
7.1 Inflectional morphology
Most morphologists working within the lexicalist framework assumed that
derivation and inflection are different morpholo
g
ical processes
.
29
I
n particular, Anderson (1982) defined inflection as the morpholo
gy
that “
is
r
elevant to the s
y
nta
x
”
: inflectional morpholo
gy
reali
z
es
all the morphos
y
ntactic
features of a word
(
Plural, Indica
t
ive, Active, etc., each specifying a
m
orphosyntactic category such as Number, Mood, Voice) depending on the
s
yntactic context in which the word is inserted. Inflection plays, therefore, the role
o
f “ad
j
usting” the words provided by the Lexicon to the morphosyntactic
r
equirements of the syntax. Anderson proposed a Weak version of Lexicalism,
c
laiming that the rules of inflectional morphology apply after syntax, intermixed
with phonolo
g
ical rules; consider the architecture of Anderson’s proposal (details
o
mitted
):
30
L
e
xi
c
al
In
se
rti
o
n
(Phonology,
Inflectional Morphology,
Semantic Interpretation)
F
i
g
ure 4
Other linguists assume that inflection and derivation are instances of the same
op
eration, i.e. affixation, but that the diff
e
r
ences between them can be ex
p
lained as a
m
atter of ‘ordering’. This approach is d
e
fended in Kiparsky’s (1982, 1983) model of
29
Cf., amon
g
others, Aronoff (1976), Scalise (1988) and Anderson (1982, 1992). The opposite view has
been also maintained however, cf. among others Halle (1973), Lieber (1980).
30
Realization-based models of morphology (e.g.
A
n
derson 1992, Stump 2001) continue in some way the
d
i
ssoc
iati
o
n
o
f th
e
infl
ec
ti
o
nal an
d
d
erivational rule types (defining
r
espectively inflect
ed wo
r
d
f
o
rm
s
v
s. lexemes
)
.
Morphology
(
Lexicon +
W
FRs
)
S
y
nta
x
T
HE LEXICALIST APPROACH TO WORD
-
F
ORMATION
175
Lexical Morphology
:
3
1
he assumes that both word-formation rules and phonological
r
ules apply in the Lexicon in an orderly progression of cycles. Without attempting to
i
llustrate the details of the theory, Kiparsky’s model can be represented as in Figure
5.
Morpholo
g
ical rules are assi
g
ned to ordered levels. Inflectional rules are
assi
g
ned to a later level than derivati
o
nal and compoundin
g
rules (thus explainin
g
why inflection usually appears outside derivation). Each morphological level is
paired with a class of specific phonological rules: the output of a word-formation
r
ule is sent to the phonological rules of that level
.
32
(
Postlexical
Phonology
,
S
emantic
L
e
xi
c
al
I
nterpretation)
I
n
se
rti
on
Fi
g
ure 5
A
s it can be seen, both Anderson’s and Kiparsky’s proposals highlight that
i
nflection is a problem for the strong lexicalist model in Figure 3. An alternative
point of view has been proposed by Booij (1996, 2002: 19-20) who claims that there
are two types of inflection:
i
nherent inflection (which “adds morphosyntactic
properties with an independent semantic value to the stem of the word”) and
contextual inflection (“required by the syntactic context, but [which] does not add
i
nformation”
)
. A functional continuum ra
ng
in
g
from strictl
y
word-formin
g
processes
(lexical) to strictl
y
contextual inflection (s
y
ntactic) can be ima
g
ined, placin
g
i
nh
e
r
e
nt infl
ec
ti
o
n
so
m
ew
h
e
r
ec
l
ose
r t
ode
ri
v
ati
o
n
.
Thi
sd
i
s
tin
c
ti
o
n
c
an a
ccou
nt f
or
t
he fact that, sometimes, inherent inflection (but never contextual inflection) may
feed WFRs: com
p
aratives and su
p
erlatives are found as
p
arts of com
p
ounds and
de
ri
ved wo
r
ds.
3
1
Developing from Siegel’s (1974) Level Ordering Hypothesis, the theory of Cyclic Phonolog
y
developed in Mascaró (1976) and Pesetsky
(
1979
)
. A number of different versions of L
exical
M
orphology/Phonology
h
ave been proposed, somewhat differing with respect to the characterization
o
f rules (morpholo
g
ical or phonolo
g
ical) or proposin
g
a
d
iff
e
r
e
nt l
eve
l-
o
r
de
r
in
g
or
g
anization (cf.
Mohanan (1986), Kiparsky (1985), Halle & Mohan
a
n
(1985), Booi
j
& Rubach (1987)). All these
approaches embrace Strong Lexicalism, but at the cost of enhancing the Lexi
c
on
(
which now, besides
W
FRs, must contain phonolo
g
ical rules).
32
The rules of lexical phonology are cyclic because th
e
y apply at each level of word-formation, before
t
he next level can take place. On the other hand, the rules of postlexical phonolo
gy
appl
y
after all
word-formation and s
y
ntactic processes have taken place (the
y
are not c
y
clic).
L
e
xi
co
n
Un
d
er
i
ve
d
i
tems
Level 1 morph. Level 1 phon.
Level 2 morph. Level 2 phon.
Level n mor
p
h. Level n
p
hon.
S
y
nta
x
1
76
S
ER
G
I
OSC
ALI
S
E
&
EMILIAN
OGU
E
V
AR
A
7.2 Syntactic Morphology
One of the key arguments against the Lexicalist Hypothesis has been Occam’s
r
azor: if it can be demonstrated that a satisfactor
y
explanation of morpholo
g
ical
phenomena is achievable without an
y
additiona
l
theoretical apparatus (i.e. onl
y
with
t
he instruments of s
y
ntax and, more rarel
y
, phonolo
gy
), then dispensin
g
with the
m
orphological component of the grammar is a desirable effect for linguistic theory.
This idea has, in one way or another, dominated the theoretical debate since the
1990’s: develo
p
ments in recent theories of
s
yntax seem (more) capable of dealing
with processes of word-formation than they were in the 1970’s.
Syntactic models of morphology argue that word-formation phenomena follow
s
yntactic constraints, interacting with syntactic operations, and that they should be
s
ubsumed within the s
y
ntactic component.
7.3 The Syntactic Incorporation Hypothesis
Incor
p
oratio
n
33
has been at the center of many
t
h
eo
r
e
ti
c
al
d
i
scuss
i
o
n
s bec
a
use
it
c
onsists of a “formally morphological process with syntactic implications” (Mithun
2000: 923-24). The central issue of this debate is whether incorporation is a lexical
o
r a syntactic process. Consider the
e
xample (38a), a prototypical case of
i
ncorporation from Classical Nahuatl and
t
he synonymous syntactic construction
(
38b
)(
Iturrioz Leza 2001: 715
)
:
b.
n
i-c-qua in naca-tl
1
S
G
.
S
U
B
J
-3
S
G
.
OB
J
-e
at
D
E
T
m
e
at-ABS
‘I eat
(
the
)
meat’
The complex verb
naca
-
qua
in (38a) is formed b
y
incorporatin
g
the noun
naca
i
nto the verb. In true incorporating languages, both the incorporated NV sequence
(
38a) and the full sentence in (38b) are semantically equivalent and grammatical.
B
aker (1988) analyzed such constructions as involving “the syntactic movement
o
f a word-level category from its base pos
i
ti
o
n t
oco
m
b
in
ew
ith an
o
th
e
r
wo
r
d
-l
eve
l
c
ategory” (Baker 1988: 424). Baker’s syntactic incorporation hypothesis i
s
developed within the framework of
Government-Binding
(Chomsky 1981): it is
g
desc
ri
bed
a
s
an in
s
tan
ce o
f h
e
a
d
-t
o
-h
e
a
d
m
ovement that leaves a trace (a properl
y
g
overned empt
y
cate
g
or
y
) with which the moved element must be co-indexed.
C
onstraints on the process are explained b
y
the
E
mpt
y
Cate
g
or
y
Principl
e
:
th
e
c
rucial
p
oint is that a trace of the incor
p
orated element must be an argument of V,
33
The best-known type of incorpora
t
ion involves adding the internal argument (i.e. direct object) into the
v
erb but many other patterns are attested (e.g. ver
b
incor
p
oration,
p
articl
e
incor
p
oration,
p
assive
i
ncor
p
oration,
p
ronoun inco
rp
oration, cf. Mithun 2000, Iturrioz Leza 2001
)
.
(
38
)
a.
n
i- naca-
q
ua
1
S
G
.
SU
BJ m
e
at-
e
at
‘I meat-eat’