?tekauer Pavol, Lieber Rochelle. Handbook of Word-Formation
Подождите немного. Документ загружается.
14
7
Štekauer, P. and R. Lieber (eds.), Handbook o
f
Word-Formation
,
14
7
—18
7.
© 2005 Springer. Printed in the Netherlands.
THE LEXICALIST APPROACH TO
WORD-FORMATION AND THE NOTION OF THE
LEXICON
1
SERGIO SCALISE AND EMILIANO G
U
EVARA
1. A DEFINITI
O
N
Th
e
t
e
r
m
L
exicalism
refers to the theoretical standpoint in modern generative
l
inguistics according to which the process
e
s
that form complex words (derivation
and compoundin
g
2
) are accounted for by a set of L
exical
R
ules
,
i
ndependent of and
d
i
ff
erent
f
rom the s
y
ntactic rules of the
g
rammar (i.e. word formation is not
performed b
y
s
y
ntactic transformations). Such Lexical Rules are assumed to operate
i
n a pres
y
ntactic component, the L
exicon
.
The Lexicalist approach to word-formation can be said to begin in the early
1970s with two fundamental articles: Chomsky’s
R
emarks on Nominalizations
(
1970
)
and Halle’s Prolegomena to a Theory of Word Formation
(
1973
)
. Since then,
l
exicalism developed in a linear and constant way with an impressi
ve se
ri
es o
f
wo
rk
s
which contributed to shape a model that – in its basic tenets – has been adopted now
for more than 30 years.
Sie
g
el (1974) desi
g
ned a level-based morpholo
g
ical model while Jackendoff
(1975) explored the relation between the formal and the semantic parts of
m
orpholo
g
ical operations b
y
means of Redundanc
y
Rules. Shortl
y
after, Aronoff
(
1976
)
established the foundations of the whole disci
p
line with the first
c
omprehensive monograph in generati
v
e morphology. Focusing on derivational
processes, Aronoff improved the notion of rule and developed an articulated syste
m
o
f restrictions in order to constrain the excessive power of
W
ord Formation Rules
(WFRs); he also envisioned the relevance of the notion of productivity and proposed
a word-based morpholo
gy
.
Man
y
specific studies were to follow, extendin
g
the Lexicalist approach to an
e
ver-
g
rowin
g
variet
y
of lan
g
ua
g
es and issues. Amon
g
them we can mention Booi
j
(1977) on Dutch, Allen (1978) on English, Pesetsky (1979) on Russian, and Scalise
(
1980
)
on Italian. Still, numerous othe
r
p
ublications introduced new fundamental
c
oncepts which contributed to build up a complete and consistent morphological
1
We have been able to write this
c
hapter also thanks to funding by the Italian Ministry of Research and
U
niversit
y
. We would like to thank Stephen And
e
r
son, Geert Booi
j
, Rochelle Lieber and Pavol
Štekauer for their comments on previous drafts. W
e
are
g
rateful to Ni
g
el James for lookin
g
over ou
r
En
g
lish. Needless to sa
y
, the authors are solel
y
re
s
ponsible for the final version of this chapter. Even
though we discussed together the whole plan of this chapter, Sergio Scalise is responsible for sections
2 through 4, and Emiliano
G
uevara, 5 through 8.
2
Generally, though some versions of lexicalism explain also inflection by means of lexical rules. Cf.
sec
ti
o
n 5
.
1
be
l
ow.
1
48
S
ER
G
I
OSC
ALI
S
E & EMILIAN
OGU
E
V
AR
A
theory: for example, Lieber (1980) proposed the mechanism of feature percolation
,
W
illiams
(
1981
)
formulated an im
p
or
t
ant generalizati
o
n on morphological
heads
,
Selkirk
(
1982
)
refined the
l
evel ordering hypothesis
,
Anderson (1982), brought
i
nflectional morphology into the picture. This list of publications is far from being
c
omprehensive or even fair to the many scholars who took part in the developments
o
f lexicalist morphology: the approach gradually developed into an articulate set of
hy
potheses, an autonomous vocabular
y
and specific anal
y
tic techniques.
Plan of the chapter: section 2 is devoted to overview the historical developments
i
n morpholo
gy
precedin
g
the lexicalist approach in
g
enerative lin
g
uistics. In section
3, we describe the characteristics of the notion of Lexicon in the early days of
g
enerative linguistics. Section 4 examines in detail the two seminal works in the
l
exicalist framework, namely Halle (1973) and Aronoff (1976). Then (section 5),
s
ome of the most important enhancements to the lexicalist approach are briefly
i
llustrated. In section 6, we return to the notion of Lexicon as it has been refined by
the lexicalist tradition. Finally (sections 7 and 8), we discuss some ma
j
or problems
(
speciall
y
the relation between morpholo
gy
and s
y
ntax) that an up-to-date lexicalist
v
i
ew
ha
s
t
oco
nfr
o
nt
w
ith
.
2. A BRIEF HI
S
T
O
RY
In the organization of linguistic theory, morphology is an intermediate discipline
between phonology and syntax. Over the years, the relationship of morphology with
these two fields of research has changed substantially a good number of times.
In 1
9
th
century European comparative linguistics, morphology lay at the heart of
the reconstruction of the Indo-European languages (cf. Bopp 1816, to name but one,
who compared Sanskrit, Latin, Persian and the Germanic lan
g
ua
g
es b
y
lookin
g
almost exclusivel
y
at morpholo
g
ical phenomena).
Morpholo
gy
was also central in American structuralist lin
g
uistics, althou
g
h the
m
ain focus of research was on phonology. The major ‘discovery’ of the time, the
P
honemic Princi
p
l
e
,
could be easily extended to morphology giving rise to the well-
k
nown
p
arallelis
m
p
hon
e
:
p
honem
e
= mor
p
h
:
mor
p
heme
,
extensively used to study
allophonic and allomorphic variation. Morphology was always present in the
A
merican tradition either viewed as grammatical process (Sapir 1921), or as
arrangement of morphemes (Bloomfield 1933)
.
3
W
ith the advent of earl
y
Generative Grammar (Chomsk
y
1957), morpholo
gy
lost
r
elevance in the
g
eneral or
g
anization of theoretical lin
g
uistics. Within that
framework, the lexicon contained onl
y
simple words (idios
y
ncratic, arbitrar
y
si
g
ns);
n
either com
p
ounds nor derived words had a place there. The only location where
they could be constructed was the transformational component, which at the time
was the only theoretical device capable of expressing grammatical relations. Phrase
s
tructure rules and transformations were allowed to manipulate words and
morphemes, thus making superfluous ever
y additional specification to account for
r
r
3
Cf. the discussion of the different models existing at the time in Hockett
(
1954
)
. Cf. also the relevance
o
f morphological matters in classical anthologies, e.g. Joos (1957).
T
HE LEXICALIST APPROACH TO WORD
-
FORMATION
149
t
h
es
tr
uc
t
u
r
eo
f
wo
r
ds.
At th
es
am
e
t
ime, all the possible variations in form th
a
t
w
ords and morphemes might show (allomorphy) were assigned to the phonological
c
omponent.
F
or instance
,
in Aspects (1965: 184) Chomsk
y
proposed to use ‘nominalization
t
ransformations’ to account for the relation between word-pairs such as
d
estro
y
/
dest
r
uction
claimin
g
that “phonolo
g
ical
ru
l
es w
ill
de
t
e
rmin
e
that
n
om+destro
y
beco
m
es
dest
r
uction
.
” Hence, the purely morphological relation
be
t
ween
destro
y
an
d
dest
r
uction
was at the time accounted for by a combination o
f
s
yntactic and phonologi
c
al o
p
erations. In addition,
i
nflectional morphology was
h
andled in a similar way: Chomsky & Halle (1968) analyzed both irregular an
d
r
egular inflected verb-forms like
sang
and
g
mended
, as
sing
+past and
g
g
mend
+past,
d
d
“
where past is a formative with an abst
r
a
c
t f
e
at
u
r
es
tr
uc
t
u
r
e
i
n
troduced by syntactic
r
ules”
(
1968: 11
)
.
A
t that time
,
4
g
enerative lin
g
uistics simpl
y
did not have adequate formal
m
echanisms for these phenomena: the theor
y
assumed no morpholo
g
ical rules at
all
.
5
However, transformations were not suited to explain morphological facts: they
h
ad been introduced to handle syntactic phenomena, i.e. totally productive,
t
ransparent and regular phenomena. Words, on the contrary, tend to be less regula
r
(
cf.
d
estro
y
ĺ
*
d
estroy-ation), and, sometimes, they undergo idiosyncratic
l
exicalization
(
cf.
t
r
ansmission
‘the action of transmitting’ vs.
t
r
ansmission
‘gearbox
o
f a car’); furthermore, most lexical processes are not full
y
productive (cf.
read
ĺ
*
read
-
ation
).
A
few
y
ears later, Chomsk
y
’s Remarks on Nominalizations
(
cf. Roe
p
er, this
v
olume) suggested that these facts could be better explained by lexical rules: “Fairly
i
diosyncratic morphological rules w
i
ll determine the phonological form of
r
efus
e
,
d
estro
y
,
etc., when these items a
pp
ea
r
i
n the noun position” (Chomsky 1970: 271).
4
A summar
y
of the situation of morpholo
g
ical theor
y
in the be
g
innin
g
s of Generative Grammar is
g
iven
i
n Anderson
(
1988: 147
)
:
“
In American structuralist terms, the enterprise
o
f morphology can be divided into the study of
m
orphotactics (the arrangement of morphological elements into larger structures) and allomorphy
(variations in the shape of the ‘same’ unit). Early generative views, typified by Chomsky (1957) or
L
ees (1960) assigned the arrangement of all items into larger constructions to the syntax, whether the
s
tr
uc
t
u
r
es
in
vo
l
ved we
r
e
a
bove o
r
be
l
ow
th
e
l
eve
l
o
f th
ewo
r
d
–
w
hi
c
h
e
f
f
ectively eliminated the
f
f
independent study of morphotactics. The program
of classical generative phonology, on the other
m
h
and, as summed up in Chomsk
y
& Halle (1968), was
t
o reduce all variation in shape of unitar
y
l
in
g
uistic elements to a common base form as thi
s
mi
g
ht be affected b
y
a set of phonolo
g
ical rules –
which effectivel
y
reduced the stud
y
of allomorph
y
to the listin
g
of arbitrar
y
suppletions. With nothin
g
o
f substance left to do in morphology, genera
t
ive linguists had to be either phonologists o
r
syntacticians.”
5
In the major ‘readings’ that made the history o
f
Generative Grammar (e.g. Fodor & Katz 1964, Bierwish
&
Heidol
p
h 1970, Jacobs & Rosenb
a
um 1970, Reibel & Shane 1969, Gross, Halle & Schuetzenberger
aa
1973) there are no more than three to four articles on morphology and, crucially, they all treat purely
m
orphological phenomena transformationally (exceptions exist, e.g. Kiefer (1973)). Another famous
r
eader (Peters 1972) had the significant title Goals of Linguistic Theor
y
(with contributions by
l
inguists such as Fillmore, Chomsky, Postal, Kipa
r
sky, etc.): the volume contains only papers on
syntax and phonology, thus showing that at the ti
m
e
morphology was not even considered among the
‘
goals’ of linguistic theory.
1
50
S
ER
G
I
OSC
ALI
S
E & EMILIAN
OGU
E
V
AR
A
Linguists gradually became convinced that
rules different from transformations
t
s
hould o
p
erate in the lexicon to form com
p
lex words
.
6
2
.1 Lees
(
1960
)
B
efore going into the details of the lexicalist approach to morphology, let us
briefl
y
consider the kind of problems that
e
arl
yg
enerative theor
y
had to face. The
m
ost exhaustive treatment of morpholo
g
ical phenomena (specificall
y
of nominal
c
ompounds) within a transformational framework is that of Lees (1960).
L
ees accounts for the formation of compound words by applying transformations
to sentences. The arguments in support of this proposal are essentially of a semantic
and syntactic nature, and can be summarized in the following three points:
(
a) Nominal compounds are generated by transformations from underlying sentence
s
tructures in which the grammatical relations (subject, object, cf. p. 119) that
h
old implicitly between the two forma
t
ives of the com
p
ound are ex
p
ressed
explicitly.
(
b) If the meaning of a compound is ambiguous, it is possible to show that this
ambiguity results from different underlying sentences corresponding to the
different meanin
g
s. For instance, the ambi
g
uit
y
of a compound such as
snake
poison can be explained in “
g
rammatical” terms b
y
derivin
g
the different
m
eanin
g
s from (at least) th
r
ee different sentences
(
a. X extracts poison
f
rom the
snake
,
b.
T
he snake has the
p
oison
,
c. The poison is for the snak
e
).
(
c
)
Transformations can account for the intuition that com
p
ounds such as
windmill
an
d
flour mill
express different grammatical relations despite their superficial
l
s
imilarity (both compounds are of the N+N type): they are derived from different
deep structures (a. Wind powers the mill
,
b.
T
he mill grinds the flour
).
T
he ma
j
or problem arising from Lees’s proposal is that it requires a good deal of
deletion transformations which are too powerful
.
7
For example, to for
m
windmill
from the underlying sentence
wind powers the mill
it is necessary to delete the verb
l
power
,
while to derive the compound
car thief
, it is necessary to delete the verb
f
f
steal
(
assumin
g
that the deep structure of the compound is the sentence the thie
f
steals the
car
). In other words, Lees’s transformations require the deletion of lin
g
uistic
material that cannot be independently said
to have been ther
e in the first
p
lace.
Subsequent developments in generative
l
inguistics made an effort to exclude
unconstrained rules such as these from the g
r
ammar. It was already clear, by the mid
1960’s – with the influence of Chomsky’s (1965) principle of
R
ecoverability of
T
ransformations – that this type of unrestricted operations could not bring us close
r
to an adequate characterization of the notion ‘natural language’. Even if it were
possible to formulate the transformations required in the derivation of compounds in
s
uch a wa
y
as to satisf
y
the recoverabilit
y
principle, it would nevertheless be
n
ecessar
y
to postulate at least as man
y
transformations as the number of verbs that
6
Cf. Wasow
(
1977
)
and the comments thereof in Anderson
(
1977
)
.
7
A detailed discussion of this topic can be
f
ound in Allen (1978) and Scalise (1984), among others.
ff
T
HE LEXICALIST APPROACH TO WORD
-
FORMATION
151
could be deleted, thus assuming rules of ‘
power
‘
deletion’, ‘
r
steal
deletion’, etc. These
l
t
ransformations are clearly
ad hoc
(
each of them is constructed to account for one
example) and their number cannot be const
r
ained; in this respect, Lees’s proposal
fails to achieve a satisfactor
y
level of descriptive adequac
y
.
3. THE LEXI
CO
N
Th
e
n
o
ti
o
n
o
f L
exicon
in Generative Grammar has
g
one throu
g
h a complex
p
rocess of develo
p
ment, from a Bloomfieldian conce
p
tion
(
the lexicon is “a list of
i
rregularities”, idiosyncratic
s
ound-meaning pairings) to a
m
ore articulated vision
,
where besides irregularities, the Lexicon contains also regular processes
.
8
In
S
yntactic Structures (Chomsky 1957), the Lexicon was not regarded as an
autonomous component of the grammar: the rules introducing lexical items were the
l
ast rules of the categorial component. Thus, the categorial component had only one
ty
pe of rule both for expandin
g
cate
g
orial s
y
mbols (e.
g
. 1.i, 1.ii) and for introducin
g
l
exical items (e.
g
. 1.v-vii):
(
1
)
i. S
ĺ
NP + Aux +
V
P
ii
.
NP
ĺ
D
e
t + N
iii.
V
P
ĺ
V
+ NP
i
v.
A
u
x
ĺ
pres., past
v.
D
e
t
.
ĺ
th
e…
v
i
.
N
ĺ
g
irl, book ...
vii.
V
ĺ
take
,
read
,
walk …
F
urthermore, only simple words (or, better, morphemes) could be inserted by
suc
h r
u
l
es.
The most important modification of the framework (as far as morpholo
gy
is
c
oncerned, and in
p
articular for the lexicalist a
pp
roach
)
was the se
p
aration of the
L
exicon from the rewritin
g
rules proposed in Aspects. This move permitted a
significant simplification of the grammar since many of the properties of lexical
formatives are, in fact, irrelevant to the functioning of the base rules and,
furthermore, are often idiosyncratic. For exam
ple, the fact that there are two classes
m
m
o
f transitive verbs, those that allow deletion of the ob
j
ect (e.g.
to read
) and those
dd
t
hat do not (e.g. to bu
y
), no longer had to be handled by rewriting rules. Instead, the
properties of verbs such as
read
and
d
b
u
y
could be specified in their lexical entries.
The or
g
anization of the lexicon presented in Aspects can be considered a ma
j
or step
i
nto the direction of what we call toda
y
Lexicalism: in order to simplif
y
the s
y
ste
m
o
f base rules, the Lexicon was
g
iven a
g
reater importance (and wei
g
ht) in the theor
y
.
I
n Chomsky’s As
p
ects
,
atomic category symbols wer
e
subcategorized by feature
m
atrices
(
sets of attribute-value
p
airs
)
i
n
w
hi
c
h
v
ari
ous
kin
ds o
f inf
o
rmati
o
n
we
r
e
encoded. Two types of feature matrices were used. The first was a system of
i
nherent features describing lexical entries (cf. (2)):
8
F
o
r a
c
l
e
ar
d
iff
e
r
e
ntiati
o
n
be
t
wee
n th
ese
t
wo
c
onceptions of the lexicon, cf. Aronoff (1989).
1
52
S
ER
G
I
OSC
ALI
S
E
&
E
MILIAN
OGU
E
V
ARA
T
he second type of subcategorization consisted in a system of contextual features
that specified the syntactic environment i
n
which a given lexical entry can appea
r
(
so-called strict subcategorization frames).
C
onsider the contextual features that
c
haracterize the following verbs:
Th
e
inf
o
rmati
o
n a
ssoc
iat
ed w
ith a
l
exical item (lexical cate
g
or
y
an
d
s
ubcate
g
orization frames) is crucial for the explanation of the morpholo
g
ical
phenomena in an
y
lan
g
ua
g
e: morpholo
g
ical rules use this information to select the
bases to which they may apply. For a br
i
ef illustration, consider the following nouns,
with each one being able to appear only w
i
th certain derivational suffixes
,
but not
w
ith all
o
f th
e
m
:
(
4
)
-ian -hood -(i)fy
C
homsky + – – = Chomskyan
boy – + – = boyhood
beauty – – + = beautify
(
3
)
eat
[+__NP ]
t
John eats food
.
seem
[+__Adjective, +__
like
-
Predicate-Nominal
]
John seems sad. John seems like a nice fellow.
grow
[+
__
NP, +
__
#, +
__
Ad
j
ective ]
John
g
rew a beard. John
g
rew. John
g
rew sad.
(
2
)
CO
MM
O
N
+ –
C
O
U
NT
A
NIMAT
E
+– +
–
A
N
IMATE AB
S
TRA
CT
HU
MAN
E
gyp
t
+
–
+ – +
–
H
U
MAN
book
vi
r
tue
di
r
t
Fr
eud
F
ido
+
–
bo
y
d
og
boy = [+common, +count, +animate, +human]
Egypt = [–common, –animate], etc.
THE LEXICALIST APPROACH TO WORD
-
FORMATION
153
Th
esu
ffix -
ian
can only attach to [–common] nouns (cf.
E
gyptian
,*
table
-
ian
,
*
ho
r
se
-
ian
)
, whereas -(i)f
y
attaches to nouns subcategorized as [+common,
±
abstract, –animate]
(
cf. countrif
y
,*
E
ngland-if
y
,*
s
ecretar-if
y
,*
hors-if
y
). Finally,
-
hood
selects [+common, –abstract, +animate] nouns (cf.
d
neighborhood
,
priesthood
,
*
democracy-hood
,*
correction-hood
, *
George-hood
).
dd
Moreover, strict subcate
g
orization frames are also relevant for derivation, as it is
o
ften the case with deverbal derivation. For instance
,
the suffix
-able
attaches b
y
r
ule to verbs characterized as [+
__
NP] (i.e. transitive verbs):
eatable
,
believable
bu
t
*
seemable
.
T
he notion of lexicon
p
resented in
A
s
p
ects was yet a static one: it served the sole
function of providing the syntax with words. Chomsky informally suggested that “it
m
ay be necessary to extend the theory of lexicon to permit some ‘internal
c
omputation’” (Chomsky 1965: 187). As we w
i
ll see, in later work, (e.g. Chomsky
1970, Halle 1973, Jackendoff 1975, Aronoff 1976
)
the Lexicon is much more than a
s
imple list of words: besides a list of underived lexical entries and idios
y
ncratic
c
omplex units, it also contains an explicit internal computation, the word formation
c
om
p
onent.
4
. LEXI
C
ALI
S
M
A
s it has been pointed out above, Chomsky’s
R
emarks on
N
ominalization
i
nitiated a totally different perspective
o
n morphological phenome
n
a, suggesting that
at least some complex words are better
e
xplained as lexical formations than as
t
ransformations: derived complex words are built in the Lexicon, inflected complex
words are
g
enerated b
y
s
y
ntactic transformations (cf. Roeper, this volume). This
i
dea laid the foundations for a more d
y
namic view of the Lexicon.
R
emarks on Nominalization represented a definitive turnin
g
point in the stud
y
of
m
orphology; its consequences greatly exceeded what could be imagined at the time.
C
homsky treated only deverbal nominalizations but, quite clearly, the lexicalist
explanation could be argued for many ot
her derivational processes: deverbal
t
ad
j
ectives (
readable
,
att
r
active
), dead
j
ectival nouns (
r
eadabilit
y
,
att
r
activeness
)
etc.
(
cf. Carstairs-McCarthy 1992: 17–19). In the following sections we will provide an
o
verview of some of the most influential developments in the lexicalist approach. In
particular, those which had the most profound impact on the architecture of the
g
rammar
.
4
.1 Halle
(
1973
)
T
he linguist who was first able to draw the logical conclusions from the
c
riticisms that were being raised against the transformational treatment of word
f
o
rmati
o
n
w
a
s
M
o
rri
s
Hall
e.
In hi
s
arti
c
l
e
P
rolegomena to a Theory of Word
F
o
r
mation
,
Halle advanced the first proposal of an autonomous morpholo
g
ical
c
omponent within the framework of the Lexicalist theor
y
. This work, althou
g
h
154
S
ER
G
I
OSC
ALI
S
E
&
EMILIAN
OGU
E
V
ARA
r
elatively short and programmatic, has ser
v
ed as the foundation and ins
p
iration fo
r
muc
h
wo
rk in l
e
xi
c
ali
s
m
.
H
alle’s starting point is the idea that if a grammar is a formal representation of a
n
ative speaker’s knowledge of language, then there must be a component that
accounts for the speaker’s lexical knowledge: a speaker of English ‘knows’, fo
r
example, a) that
read
is a word of his/her language, but
d
le
z
en
is not, b
)
that certain
w
ords have internal structure (e.
g
.
un
-
d
r
ink
-
able
)
and c
)
that the internal structure
r
espects a specific order of c
o
ncatenation of morphemes (
un
-
d
r
ink
-
able
is a possible
wo
r
dbu
t
*
un-able-drink
or *
k
d
r
ink
-
un
-
able
are not). Such knowled
g
e can be
r
e
p
resented in a formal model:
F
i
g
ure 1
Let us briefl
y
examine how this model works. In Halle’s view, the basic units of
t
he lexicon are mor
p
hemes. Each mor
p
heme is re
p
resented as a se
q
uence of
phonological segments provided with a syntactic label (N, V, etc.), except affixes,
which are labeled Af (without syntactic category):
(
5
)
[home]
N
[
arrive
]
V
[
-al
]
A
f
These forms constitute the in
p
ut of
W
ord Formation Rules
,
which are
responsible for the sequential arrangement
of the morphemes of a language into
t
actual words. WFRs can freely apply to the list of morphemes, forming potential
words (i.e. words that are not listed in the Dictionary). WFRs are of two types, those
t
hat apply to stems and those that apply to words:
(
6
)
a. [STEM + Af ]
A
t
o
t+
al
b. [VERB + Af
]
N
arri
v
+
al
W
FRs can change not only the lexical category of the input lexical item, but also
t
he s
y
ntactic features associated with it. For example, the suffix
-hood
can attach to
d
[
–abstract] Nouns resultin
g
i
n [+abstract] Nouns
(
b
o
y
ĺ
boyhood
,
p
ries
t
ĺ
priesthood
).
d
d
L
i
s
t
of
Mor
p
hemes
W
ord
Fo
rmati
on
R
u
l
es
Filt
e
r
Di
ct
i
onary
o
f
W
ords
O
ut
p
ut Phonolog
y
Synta
x
THE LEXICALIST APPROACH TO WOR
D
-
F
ORMATION
155
Halle’s model of morphology entails a more radical version of the L
exicalist
H
ypothesis than was originally proposed in Chomsky (1970): Halle’s WFRs operate
i
n the same way for derivation and inflectio
n
9
(compounding was not yet in the
picture at the time), both derivational and inflectional affixes are equall
y
listed in the
L
exicon. The motivation for this choice is that the set of morphophonolo
g
ical
operations and the idiosyncratic behavior
typical of derivational processes has
r
p
arallels in inflection.
The third subcom
p
onent in Halle’s model is the F
ilte
r
.
The Filter has basically
t
wo functions: a) it adds idiosyncratic features when necessary (e.g. a word such as
recital
is formed regularly as is
l
a
rr
ival
,
but the Filter gives it its idiosyncratic
m
eaning “performance of a soloist”) and b) it prevents possible but non-existing
words (e.g.
*
i
gnoration) from appearing in sentences by assigning the feature [
–
l
exical insertion
]
to them.
The final subcom
p
onent in Halle’s model is the
D
ictionar
y
,
which is entirel
y
determined b
y
the interaction of the List of Morphemes, WFRs and the Filter: it
c
ontains all existing words, including all
t
h
e
infl
ec
t
ed
f
o
rm
s
o
f every word. The
w
h
o
l
e
m
ec
hani
s
m
c
an
be su
mmarized with the following schema:
List of morphemes WFRs Filter Dictionary
Figure 2
Halle’s proposal represented a fundamental step in the development of the theory
o
f morphology: for the first time it was proposed to handle all morphological
phenomena in a single place (i.e. the Lexi
c
on) and by means of specific rules
(WFRs). This suggestion served not only to lighten the burden of the
t
ransformational component b
y
removin
g
t
hose operations that involved numerous
l
exicall
yg
overned exceptions: it also provided a wa
y
to account for a fundamental
difference between syntax and morphology. The notion “possible but non-existent
9
Halle’s position is today called
S
trong Lexicalist Hypothesis as o
pp
osed to the so-called Weak Lexicalis
t
Hy
pothesis
(
where derivational and inflec
t
ional morpholo
gy
are handled b
y
different sets of rules).
C
f. section 5.1 below.
1.
f
r
i
en
d
2. boy
-h
ood
3
.
r
ec
it
e
-al
4
. i
g
nore
-ati
o
n
5.
m
ou
ntain
-
a
l
X
[
+idios
y
n]
X
[– L. I.]
156
S
ER
G
I
OSC
ALI
S
E
&
EMILIAN
OGU
E
V
ARA
word” became crucial for morphology; a parall
e
l
concept is totally absent in syntax,
where in fact, it makes no sense to say that a sentence is ‘possible but non-existent’.
W
ith respect to earlier models of morphology, such as the
I
tem and Arrangemen
t
mode
l
1
0
that
j
oined morphemes by the simple operation of concatenation, Halle’s
proposal represented a significant innovation in that it contains a special mechanis
m
for creating complex words (i.e. WFRs), a m
ec
hani
s
m that mak
es use o
f m
o
r
e
l
in
g
uistic information and carries out more abstract operations than simple
co
n
c
at
e
nati
o
n
.
To see wh
y
concatenation is not an ad
e
quate wa
y
to explain word formation,
co
n
s
i
de
r th
e
Fr
e
n
c
h
wo
r
d
rest
r
uctu
r
ation
‘restructuring’: this word cannot simply be
built up by concatenation of the three morphemes
re
+
st
r
uctu
r+
ation
;
it must have
i
nternal structure since it is not
p
ossible to
attach just any pref
ix or suffix to any
f
f
base. That is, the prefix
re-
must be attached to verbs, not to nouns
(
cf.
*
re
-
ve
r
ité
‘re-truth’) or adjectives (cf.
*
re-grand
‘re-great’), and the suffix
d
-
ation
m
us
t al
so be
attached to verbs, not to nouns
(
cf.
*
v
erity-ation ‘truth-ation’) or adjectives (cf.
*
g
rand-ation ‘
g
reat-ation’). Thus the base of the derived word
rest
r
uctu
r
ation
m
us
t
be a verb
(
s
tructur
(
er
)
‘to structure’
)
and the derivation must be carried out in two
s
te
p
s: i.
s
tructur
(
er
)
ĺ
r
e-structur
(
er
)
,
ii.
r
estructur
(
er
)
ĺ
r
est
r
uctu
r-
ation
.
Th
us
the word in question has the following internal structure:
(
7
)
[[re+[structur]
V
]
V
+ation
]
N
I
t should be remembered that Halle’s main
p
ur
p
ose was to stimulate discussion
i
n a still neglected field. We can see today that this goal has, in fact, been completely
achieved although later research on morphology has raised questions on every
subpart of the model seen above.
Placing morphemes at the base of the s
y
s
tem is a problematic choice because,
while in English simple words and morphemes coincide most of the time, this is not
alwa
y
s the case with other lan
g
ua
g
es. Moreover, considerin
g
derivational and
i
nflectional affixes as included in the List of Morphemes obscures the difference
between the formation of new words (or ‘lexemes’, e.
g
.
w
r
it
+
er
from
r
w
r
ite
)
and the
formation of word-forms (e.g.
w
r
ite
+
s
,
writ+in
g
).
Halle’s WFRs are
q
uite unrestricted; no
t
only do they have access to information
t
contained in later stages in a derivation (t
he Dictionary), but they also generate a
t
l
arge number of ungrammatical forms
.
Finally, also the Filter has been criticized, mainly because it is not a finite
m
echanism. The set of possible but non-existent words is not finite in the sense that
there are no
g
rammatical principles restrictin
g
the de
g
ree of complexit
y
of derived
and compound words, with the likel
y
exception of performance considerations such
as memor
y
(cf. Booi
j
1977).
10
Cf. Hocket
(
1954
)
.