?tekauer Pavol, Lieber Rochelle. Handbook of Word-Formation
Подождите немного. Документ загружается.
HANDB
OO
K
O
F
WO
R
D
-
F
O
RMATI
O
N
3
(
Chica
g
o: Universit
y
of Chica
g
o Press 1992), and An Inte
g
rated Theor
y
o
f
Autose
g
mental Processes
(
New York: SUNY Press 1987
)
, as well as numerous
articles on various aspects of word f
o
rmati
o
n an
d
th
e
int
e
rfa
ces be
t
wee
n
m
orphology and syntax, and morphology and phonology.
A
d
Neeleman
i
s Reader in Lin
g
uistics at Universit
y
Colle
g
e London.
Hi
s
main r
ese
ar
c
h int
e
r
es
t
s
ar
ec
a
se
th
eo
r
y, the syntactic encoding of thematic
dependencies, and the interaction b
e
t
ween syntax and syntax-external
sy
stems. His main publications include Com
p
lex Predicates
(
1993
)
, F
lexible
Sy
nta
x
(
1999, with Fred Weerman
)
,
B
e
y
ond Morpholo
gy
(
Oxford: Oxford
U
niversit
y
Press 2004, with Peter Ackema), as well as articles in Lin
g
uistic Inquir
y
,
Natural Language and Linguistic Theor
y
,
and
Y
earbook of Morpholog
y
.
F
ranz
Rainer
i
s Professor of Romance languages at the Vienna University
o
f E
co
n
o
mi
cs
an
d
B
us
in
ess
A
d
mini
s
trati
o
n
.
H
e
i
s
th
e
a
u
th
o
r
o
f S
p
anische
Wortbildungslehre (Tübingen: Niemeyer, 1993) and co-editor (with Maria
Grossmann
)
of La formazione delle parole in italian
o
(Tübingen: Niemeyer, 2004),
both of these publications being comprehensive treatments of the word-formation in
t
he respective languages.
To
m
Roeper
,
Professor of Linguistics at the University of Massachusetts,
h
as written widel
y
on morpholo
gy
and la
ng
ua
g
e acquisiton,
i
ncludin
g
compounds,
n
ominalizations, implicit ar
g
uments, and derivationial morpholo
gy
. In the field of
l
anguage aquisition, he is also Mana
g
ing Editor of Studies in Theoretical
Psycholinguistics (Kluwer), a Founding editor of Language Acquisition
(
Erlbaum
)
,
an
d
al
so
th
e
a
u
th
o
r
o
f Understanding and Producing Speech
(
London: Fontana,
1983, co-authored with Ed Matthei
)
,
Parameter Setting
(Dordrecht: Reidel, 1987,
g
with E. Williams
)
,
T
heoretical Issues in Language Acquisition
(
Hillsdale: Erlbaum,
1992, with H. Goodluck and J. Weissenborn), and the forthcomin
g
T
he Prism o
f
G
rammar
(
MIT Press
).
S
er
gio
Scalise is Professor of General Linguistics at the University of
B
olo
g
na. He is the editor of the
j
ournal Lin
g
ue e Lin
g
ua
gg
i
o
.
His pulications include
Generative Morpholog
y
(
Dordrecht: Foris, 1984
)
,
M
orfologia (Bologna: Il Mulino,
(
Cambrid
g
e: Cambrid
g
e Universit
y
Press 2004), Deconstructin
g
Morpholo
gy
1994
)
, and Le lin
g
ue e il Lin
g
ua
gg
i
o
(Bolo
g
na: Il Mulino, 2001 (with Gior
g
io
Graffi
)).
A
n
d
r
e
w
Spencer is Professor of Linguistics in the Department of
L
an
g
ua
g
e and Lin
g
uistics at the Universit
y
o
f E
sse
x
.
H
e
ha
swo
rk
ed o
n
v
ari
ous
problems of phonolo
g
ical and mo
r
pholo
g
ical theor
y
. In addition to En
g
lish, his
m
ajor language area is Slavic. He is the author of
M
orphological Theor
y
(
Oxford:
B
lackwells, 1991) and co-editor (with Arnold Zwicky) of the
H
andbook o
f
M
orpholog
y
(
Oxford: Blackwells, 1998
)
.
CO
NTRIB
U
T
O
R
S
4
W
ord-Formation (Amsterdam/Philadelphia: John Benjamins, 1998)), and
E
nglish
Word-Formation. A History of Research (1960-1995
)
.
Tübingen: Gunter Narr,
2000), and the forthcoming
M
eaning Predictability in Word-Formation
(
Amsterdam/Philadelphia: John Ben
j
amins)
Gregory T
.
Stump is Professor of En
g
lish and Lin
g
uistics at the
U
niversity of Kentucky. His research ha
s
focused on the development of Paradig
m
Function Morphology. He is the author of
T
he Semantic Variability of Absolute
Constructions
(
Dordrecht: Reidel, 1985
)
,
I
n
f
lectional Morpholo
gy
: A Theor
y
o
f
Paradi
g
m Structur
e
(Cambrid
g
e: CUP, 2001). He is currentl
y
servin
g
as an
A
ssociate Editor of Lan
g
ua
g
e a
n
d as a Consultin
g
Editor fo
r
Y
earbook o
f
M
orpholog
y
.
Bo
g
da
n
Szymanek is Professor of English linguistics, Head of the
Department of Modern English, Catholic University of Lublin, Poland. His ma
j
o
r
r
esearch interests include morphology a
n
d its interfaces with other grammatical
c
omponents, lexicology, English and Slavic languages. He is the author of
Categories and categorization in morpholog
y
(
RW KUL Lublin, 1988
)
and
I
ntroduction to morpholo
g
ical anal
y
sis
(
PWN Warsaw, 1998
(3
rd
ed.)).
d
Davi
d
Tuggy has worked in Mexico with the
S
ummer Institute of
Linguistics since 1970. His main areas of
interest include Nahuatl, Cognitive
f
g
rammar, translation, lexico
g
raph
y
, and inadvertent blends and
o
ther bloopers. He is
an a
u
th
o
r
o
f The transitivity-related morpholog
y
of Tetelcingo Náhuatl; An
exploration in Space grammar
(UCSD Doctoral dissertation, 1981), “The affix-stem
r
distinction; A Cognitive grammar analysis of data from Orizaba Nahuatl” (Cognitive
Linguistics 3/3, 237-300), “The thing is is that people talk that way. The question is
i
s why?” (In: E. Casad (ed.). 1995. Cognitive linguistics in the redwoods; the
e
xpansion o
f
a new paradi
g
m in lin
g
uistics
.
Berlin: Mouton de Gru
y
ter, 713-752.),
and “
Abrelatas
“
an
d
scarecrow
nouns: Exocentric verb-noun compounds as
i
llustrations of basic principl
es of Cognitive grammar” (
International Journal of
(
(
E
nglish Studies
(
2004
)
III, 25-61
).
Ma
r
k
Volpe is a Ph.D candidate at SUNY at Stony Brook expecting to
defend his dissertation on Japanese morphology in early spring 2005. He is currently
a visitin
g
lecturer in the Department of Humanities at Mie National Universit
y
in
T
su, Japan. He has published independently in Lingua an
d
Sni
pp
ets an
d
ha
sco
-
authored with Paolo Ac
q
uaviva, Mark Aronoff and Robert Beard.
Pav
ol
Štekauer is Professor of En
g
lish lin
g
uistics in the Department of
British and American Studies, Prešov University, Slovakia. His research has focused
o
n an onomasiological approach to word-formation and on the history of research
i
nt
owo
r
d
-f
o
rmati
o
n
.
H
e
i
s
th
e
a
u
th
o
r
o
f A Theory of Conversion in English
(Frankfurt am Main: Peter Lan
g
, 1996), An Onomasiolo
g
ical Theor
y
o
f
En
g
lish
5
Š
tekauer P. and R. Lieber (eds.), Handbook of Word-Formation, 5—23
.
©
2005
S
pringer. Printed in the Netherlands.
BASIC TERMINOLOGY
ANDREW CARSTAIRS-MCCARTHY
1. THE NOTION OF THE LING
U
ISTIC SIGN
I
n this introductory chapter I will discuss the notions ‘morpheme’ and ‘sign’ in
r
elation to word-formation. The starting-point will be Ferdinand de Saussure’s
n
otion ‘sign’ (
s
ign
e
) (Saussure 1973), which since the early twentieth century has
i
nfluenced enormousl
y
how lin
g
uists have anal
y
sed words and parts of words as
g
rammatical units. There will be no tid
y
conc
l
usion, partl
y
because Saussure himself
was va
g
ue on crucial points, and partl
y
because amon
g
contemporar
y
lin
g
uistic
theorists there is little ag
r
eement about even the most fundamental aspects of how
word-formation should be analysed and what terminology should be used in
describing it. But I hope that this chapter w
i
ll al
e
rt r
e
a
de
r
s
t
oso
m
e
o
f th
e
main ri
s
k
s
o
f misunderstanding that they are sure to encounter later
.
1
A
handbook of English syntax in the twenty-first century would not be likely to
begin with a discussion of Saussure. Why then does it make sense for a handbook on
wo
r
d
-f
o
rmati
o
n t
odoso?
Th
e
r
e
ar
e
t
wo
r
e
a
so
n
s
.
The first is that s
y
ntax is centrall
y
c
oncerned not with individual si
g
ns in Sau
ssu
r
e
’
sse
n
se bu
t
w
ith
co
m
b
inati
o
n
so
f
s
i
g
ns. That makes it sound as if word-formation, b
y
contrast, is concerned not with
c
ombinations of signs but only with individual signs. As to whether that implication
i
s attractive or not, readers can in due course form their own opinions. For the
present, it is enough to say that, in the opinion of most but not all linguists, the way
i
n which meaningful elements are combined in syntax is different from how they are
c
ombined in complex words. The second reason has to do with Saussure’s
distinction between lan
g
ua
g
e as social convention (
l
an
g
u
e
) and lan
g
ua
g
e as
utterance
(
parole
(
(
). Each lan
g
ua
g
e as
l
an
g
u
e
belon
g
s to a communit
y
of speakers
and
,
because it is a social convention
,
individuals have no control over it. On the
o
ther hand, language as parole is something that individual speakers have control
o
ver; it consists of the use that individuals freely make of thei
r
l
angu
e
in th
e
s
entences and phrases that they utter. He
n
c
e, because syntax i
s
co
n
ce
rn
ed w
ith th
e
s
tructure of sentences and phrases, Saussure seems to have considered the study of
s
yntax as belonging to the study of parol
e
,
not
l
angue (the exception being those
s
entences or phrases that are idioms or clichés and which therefore belong to
l
angu
e
because the
y
are conventional
r
ather than freel
y
construc
t
ed
)
. So, because his focus
w
a
so
n
l
an
g
u
e
rath
e
r than
p
arol
e
,
Saussure had little to sa
y
about s
y
ntax.
1
I will use ‘
S
aussure’ in this c
h
apter as shorthand for ‘Saussure’s view as presented in the
C
ours de
l
inguistique générale’
.
Th
e
Cours is a posthumous compilation based
o
n n
o
t
es o
f
v
ari
ous se
ri
es o
f
l
ectures that Saussure delivered over a numbe
r
o
f years. Apparent inconsistencies in the
C
ours may be
d
ue to developments in Saussure’s thinking ov
e
r
time or faulty note-taking on the part of the
c
ompilers or both. Nevertheless, it is the
C
ours as a whole that has influenced subsequent linguists,
an
do
n that
b
a
s
i
s
it i
s
fair t
od
i
scuss
it a
s
if it
we
r
e created by one author as a single coherent work.
6
A
NDRE
WC
AR
S
TAIR
S
-
M
CC
ARTHY
Saussure introduced his notion ‘sign’ with a famous example: a diagra
m
c
onsisting of an ellipse, the upper half c
o
ntaining a picture of a tree and the lower
h
alf containing the Latin word
a
r
bor
‘tree’ (Saussure
r
C
ours
,p
art 1, cha
p
ter 1; 99;
67
).
2
The upper half of the diagram is meant to represent a concept, or what the sign
s
ignifies (its
s
ignifié), while the lower half represents the unit of expression in Latin
that signifies it (the
s
ignifian
t
). As Saussure acknowledges, the term ‘sign’ in its
n
ormal usa
g
e seems closer to the
signifiant
than the
t
s
i
g
ni
f
ié
,
and at first one is
i
nclined to ask what the point is in distin
g
uishin
g
the
signifiant
from the sign as a
t
w
hole. Saussure’s answer li
e
s
lar
g
el
y
in his view of how si
g
ns are related to each
o
ther. Signs (he says) do not function in
i
solation but rather have a ‘value’
(
valeu
r
)
as part of a system (part 2, chapter 4; 155-69; 110-20). Concepts (signifiés
)
do not
exist in the world indepently of language but only as components of the signs to
which they belong. By this Saussure does not mean that (for example) trees have no
r
eal existence apart from language, but rathe
r
that th
e
t
e
rm f
o
r the concept ‘tree’ will
d
iff
e
r in
valeur
from one language to another de
r
pending on whether or not that
l
an
g
ua
g
e has, for example, contrastin
g
terms for the concept ‘bush’ (a small tree) o
r
t
he concept ‘timber’ (wood from trees for use in buildin
g
or furniture-makin
g
).
3
Ea
ch
signifié
has a wider or narrower scope, accord
i
n
g
to how few or how man
y
are
the related signs that its sign contrasts with. And with
s
ignifiants
,
too
,
what matters
m
ost is not the sounds or letters that compose them but their role in distinguishing
o
ne sign from another. Thus the Attic Greek verb forms
é
phe:n ‘I was saying’ and
éste
:
n
‘I stood’ both have the same structure (a prefix
e
-
,
a root
,
and a suffix -
n
)
, but
t
h
e
i
r
valeur
within their respective verbal paradigms is different:
r
éphe:n i
s
an
‘imperfect’ tense form while
éste
:
n
i
s
‘a
o
ri
s
t’
.
So far, so
g
ood, perhaps. The Latin word
a
r
bor
and the English word
r
t
r
ee
ar
e
s
imple words, not anal
y
sable into sma
l
ler meanin
g
ful parts, and each is in
Saussure’s terms a sign. But consider the word
u
nhelpfulness
,
which seems clearly
t
o consist of four elements
,
un-
,
h
el
p
,
-ful
and
l
-
ness
,
each of which contributes in a
transparent way to the
m
e
aning of the whole. Consider also the words L
ondone
r
,
Muscovite
,
Pa
r
isian
,
Roman
,
and
V
iennes
e
, all meaning ‘inhabitant of ...’, and all
c
onsisting of a stem followed by a suffix. What things count as signs here: the whole
words, or the elements composing them, or both? It is at this point that Saussure’s
exposition becomes frustratin
g
l
y
unclea
r
,
as I will demonstrate presentl
y
.
Let us call these elements ‘morphemes’. This is consistent with the usa
g
e of
B
audouin de Courtena
y
, the inventor of the term, who speaks of ‘the unification of
t
he concepts of root, affix, prefix, ending, and the like under the common term,
m
orpheme’ (Baudouin de Courtenay 1972: 151)
a
n
d defines it as ‘that
p
art of a word
which is endowed with psychological autonomy and is for the very same reason not
2
Because readers are likely to have access to Saussure’s Cours in
v
ari
ous d
iff
e
r
e
nt
ed
iti
o
n
s
an
d
translations, I will give first a reference to the releva
n
t part and chapter, then a page reference to the
1973 edition by Tullio de Mauro, and finally a page reference to the 1983 translation by Roy Harris. I
quote passa
g
es from the
C
ours in the translation b
y
Harris. I use Saussure’s ori
g
inal technical terms
l
angu
e
,
parol
e
,
signifiant
and
t
s
ignifié
,
for which no consistent En
g
l
ish equivalents have become
es
ta
b
li
s
h
ed.
3
This illustration is mine
,
not Saussure’s
,
but is in the spirit of Saussure’s discussion of how two English
wo
r
ds
s
hee
p
an
d
mutton
corres
p
ond to one French word
mouton
.
B
ASIC TERMINOLOGY
7
further divisible’
(
1972: 153
)
. It is also consistent
w
ith rough-and-ready definitions
o
f the kind offered in introductory linguistics courses, where morphemes are
c
haracterised as individually
m
eaningful units which are minimal in the sense that
t
he
y
are not divisible into smaller meanin
g
ful units
.
4
The question
j
ust posed now
becomes: Do morphemes count as si
g
ns, or do onl
y
words count, or both? Much of
t
he diver
g
ence in how the term ‘morpheme’ is used can be seen as due to implicit o
r
explicit attempts to treat morphemes as signs, despite the difficulties that quickly
ari
se w
h
e
n
o
n
e does so.
Th
ese
ar
ed
iffi
cu
lti
es
t
hat Saussure never confronts
,
because
t
he term ‘mor
p
heme’ never a
pp
ears in the
C
ours
.
In Saussure’s defence
,
one can
fairly plead that he could not be expected to cover every aspect of his notion of the
sign in introductory lectures. Yet the question that I have just posed about
m
orphemes is one that naturally arises almost as soon as the notion of the sign is
i
ntr
oduced.
A
case can be made for attributin
g
to Saussure two diametricall
y
opposed
positions relatin
g
to the role of si
g
ns in word-formation. I will call these the
m
orpheme-as-sign position and the word-as-sign position. I will first present
ev
i
de
n
ce
fr
o
m th
e
C
ours for morphemes as signs, then
p
resent evidence for words as
signs.
1
.1 Evidence
f
or the morpheme-as-si
g
n position in Saussure’s Cours
Th
ed
i
s
tin
c
ti
o
n
be
t
wee
n
l
an
g
ue an
d
parol
e
is far from the onl
y
important binar
y
d
istinction introduced b
y
Saussure in his Cours
.
An
o
th
e
r i
s
th
ed
i
s
tin
c
ti
o
n
be
t
wee
n
syntagmatic relationships (i
n
volving elements in linear succession) and associative
r
elationships (involving elements that contrast on a dimension of choice)
.
5
Elements that can be related syntagmatically include signs, and in particular the
s
ignifiants of signs, which are ‘presented one after another’ so as to ‘form a chain’
(part 1, chapter 1, section 3; 103; 70). Chains of items that form syntagmatically
r
elated combinations are called syntagmas (
s
yn
t
agmes) (part 2, chapter 5; 170-5;
121-5). Some s
y
nta
g
mas have meanin
g
s that are conventionalised or idiomatic. This
c
onventionalisation renders them
p
art of
l
an
g
u
e
. An exam
p
le is the
p
hrase
p
rendre
la mouche
(literall
y
‘to take the fl
y
’), which means ‘to take offence’ (part 2, chapte
r
5, section 2; 172; 123). However, the great
m
ajority of phrases and sentences have
m
eanings that are transparent, not idiomatic. As such, they belong to
p
arol
e
,
not to
l
angu
e
.
As examples of syntagmas that belong to
p
arol
e
,
Saussure cites
cont
r
e tous
‘against all’,
la vie humaine
‘human life’
,
D
ieu est bon
‘God is good’, and
s
’il fai
t
b
eau temps, nous sortirons ‘if it’s fine, we’ll go out’ (part 2, chapter 5, section 1;
170; 121). These phrases and sentences do no
t
constitute signs as wholes; rather,
t
4
Thi
s
r
ese
m
b
l
es
Bl
oo
mfi
e
l
d
’
sc
la
ss
i
cde
finiti
o
n
:
‘a linguistic form which bears no partial phonetic-
s
emantic resemblance to an
y
other form’ (1933: 161). One implication of the specification ‘partial’ is
t
hat two morphemes ma
y
displa
y
total phonetic ide
n
t
it
y
(so as to be homon
y
ms) or total semantic
i
dentit
y
(so as to be s
y
non
y
ms).
5
In the technical terminology of linguistics, the term ‘paradigmatic’, promoted by Louis Hjelmslev
(
1961), has come to replace ‘associative’ as the c
o
unterpart of ‘s
y
nta
g
matic’. But I will stick to
S
aussure’s term i
n
this chapter.
8
ANDRE
WC
AR
S
TAIR
S
-
M
CC
ARTHY
t
hey are made up of smaller signs, namely the words or idiomatic expressions that
t
hey contain.
O
n this basis, the question ‘Do morphemes count as signs?’ can be refined as
‘Can morphemes as such compose syntagmas that belong to parol
e
rath
e
r than t
o
l
angu
e
?
’ At first sight, the answer is yes. In the very same passage where Saussure
g
ives the examples just quoted, he cites the word
re
-
li
r
e
‘to read again’. Saussure
u
ses the h
y
phen to draw attention to the divisibilit
y
of this word into two elements,
re-
‘a
g
ain’ and
li
r
e
‘t
o
r
e
a
d
’
.
Th
ewo
r
d
reli
r
e
thus has a meanin
g
that is as
t
rans
p
arent as that of unhelp
f
ulness. Here
,
at least
,
it seems clear that Saussure
i
ntends us to analyse the morpheme
re-
as a sign, forming part of a syntagma that
belongs to parole rath
e
r than t
o
l
angu
e
.
Further evidence for this ‘morpheme-as-sign’ position seems to be supplied by
S
aussure’s discussion of suffixes such as -
ment
and
t
-
eux
,
and of zero signs. The
wo
r
ds
enseignement
‘instruction’,
t
enseigner
‘to teach’ and
r
enseignons ‘
we
t
e
a
c
h’
c
learly share what Saussure calls a ‘c
o
mmon element’. Similarly, the suffixes
-ment
an
d
-
eux
ar
e
‘
co
mm
o
n
e
l
e
m
e
nt
s
’ in th
ese
t
o
f
wo
r
ds
ensei
g
nemen
t
,
a
r
mement
‘armam
e
nt’ an
d
changement
‘change (noun)’, and in the set
t
dési
r-
eux
‘
des
ir
ous
’
(
fro
m
désir
‘desire’),
r
chaleu
r-
eux
‘warm’
(
fro
m
chaleur
‘warmth’), and
r
peur
-
eux
‘fearful’
(
fro
m
peur
‘fear’) (part 2, chapter 5, section 3; 173-5; 123-5).
r
6
Th
ese
c
ommon elements are morphemes, in ter
m
s
of our rough-and-ready definition. Are
t
hey also signs, in Saussure’s sense? Sau
s
sure hints at the answer ‘yes’ when he
d
i
scusses
a
se
t
o
f in
s
tan
ces w
h
e
r
e ove
rt
su
ffixes contrast with zero. In Czech
,
the
n
oun
žena
‘woman’ illustrates a widespread pattern in which the genitive plural
f
o
r
m
žen
i
sd
iff
e
r
e
ntiat
ed
fr
o
m th
eo
th
e
r
c
a
se
-number forms
,
such as the accusative
s
in
g
ula
r
ženu
and the nominative
p
lural
ž
en
y
,
simpl
y
b
y
the absence of a suffix. Here
t
he
g
enitive plural has as its exponent ‘zero’ or ‘the si
g
n zero’ (part 1, chapter 3,
s
ection 3; 123-4; 86). Surely then (one i
s
inclined to think
)
the accusative singular
su
ffi
x
-
u
and the nominative plural suffix -
y
,
both being morphemes in our sense,
m
ust have at least as much right as zero has to count as signs.
It is tempting to conclude that, in com
p
l
ex words, Saussure recognises individual
m
orphemes as signs provided that the complex word is regularly formed and
s
emantically transpare
n
t.
A r
e
a
de
r
o
f th
e
C
ours who looks for explicit confirmation
o
f this temptin
g
conclusion will be frustrated, however. Man
y
complex words othe
r
t
h
a
n
re
-
li
r
e
an
d
f
o
rm
so
f
žena
are discussed, but alwa
y
s it is in contexts that
e
m
p
hasise the associative relationshi
p
s of
t
he word as a whole
,
rath
e
r than th
e
s
yntagmatic relationship between the morphemes that compose it. These discussions
point away from morphemes as signs and towards words as signs, therefore.
1.2 Evidence for the word-as-sign position in Saussure’s Cours
C
losely parallel in structure to
reli
r
e
i
s
th
eve
r
b
d
é-fair
e
‘to undo’
,
also discussed
b
y
Saussure (part 2, chapter 6, section 2; 177-8; 127-8). A
g
ain he uses a h
y
phen to
draw attention to its internal structure. The meanin
g
of
d
é
f
air
e
,
at least in man
y
6
The inconsistenc
y
in the use of h
y
phens here is Saussure’s.
B
ASIC TERMINOLOGY
9
c
ontexts, seems
j
ust as transparent as that of
reli
r
e
, on the basis of the meanings of
fair
e
‘t
odo
’ an
d
dé-
implying reversal. Indeed, Saussure draws our attention to this
t
ransparency by citing the parallel formations
décoller
‘to unstick’,
r
déplacer
‘t
o
r
r
emove’ (literall
y
‘to un-place’) and
découd
r
e
‘to unsew’. However, comparin
g
the
d
i
scuss
i
o
n
o
f
reli
r
e
,
we find an important difference in emphasis here. With
reli
r
e
,
t
he emphasis was on s
y
nta
g
matic relationships. With
d
é
f
aire
,
however, the emphasis
i
s on the associative relationships that it enters into: not just with
décolle
r
,
d
é
p
lacer
an
d
découd
r
e
bu
t al
so w
ith fair
e
itself
,
r
efaire ‘to redo’
,
and contrefair
e
‘t
o
c
ari
c
at
u
r
e
’
.
Now
,
it is clear that contrefair
e
is something of an outsid
e
r
in this list
,
because its
m
eaning cannot be predicted from that of its elements faire an
d
cont
r
e
‘against’.
One might therefore have expected Saussur
e
to say something li
k
e
thi
s:
“B
ec
a
use of
i
ts unpredictable meanin
g
, the s
y
nta
g
ma contre
f
air
e
is conventionalised and belon
g
s
as a unitar
y
si
g
n to
l
an
g
ue
,
so that
cont
r
e
an
d
f
air
e
do not count as si
g
ns in this
c
ontext. However, the meanin
g
s of the other complex words I have cited are
predictable, so they are examples of syntagmas that belong to
p
arol
e
,
and in the
m
t
he mor
p
hemes
re-
an
d
dé-
,
as well as the verb ste
m
s
that accompany them, are
s
igns.” But what Saussure actually says is almost the opposite of that. The word
d
éfair
e
is decomposable into ‘smaller units’, he says, only to the extent that is
‘
surrounded by’ those other forms (
décolle
r
,
r
efaire and so on
)
on the axis of
association. Moreover
,
a word such as
dési
r
eux
is ‘a product, a combination of
i
nterdependent elements, their value [i.e.
valeu
r
]
derivin
g
solel
y
from their mutual
c
ontributions within a lar
g
e
r
unit’ (part 2, chapter 6, section 1; 176; 126). Recall that
valeur
is a property of signs, dependent on th
r
e
ir place within the sign system as a
whole. Saussure’s words here imply, therefore, that in
dési
r
eux
,
the ‘smaller unit’ o
r
‘e
l
e
m
e
nt’ -
eux
,
though clearly identifiable, is not a sign. Saussure hints that even the
roo
t
dési
r
,
in the context of this word, does not count as a sign either, although it
c
learly does so when it appears as a word on its own.
W
e are thus left with a
co
ntra
d
i
c
ti
o
n
.
Th
ewo
r
d
reli
r
e
i
sc
it
ed
in a
co
nt
e
xt that in
v
it
es us
t
o
tr
e
at it a
s
a
u
nit
of
parol
e
,
not
l
an
g
u
e
, composed of si
g
ns,
j
ust like the sentence I
f
it’s
f
ine, we’ll
g
o
out
.
On the other hand, the discussion surroundin
g
d
é
f
air
e
in
s
i
s
t
so
n it
ss
tat
us
a
s
a
u
nit
o
f
l
an
g
ue
,
a si
g
n as a whole, composed of ‘elements’ or ‘smaller units’ that are
not signs.
O
n the basis of my presentation so far, the evidence for the two
p
ositions
(morpheme-as-sign and word-as-sign) may seem fairly evenly balanced. But there
ar
eso
li
d
r
e
a
so
n
s
t
o
think that th
ewo
r
d-as-sign position more closely reflects
S
aussure’s true view.
C
onsider the French number word
dix-neuf
‘nineteen’ (literally
f
‘ten-nine’). In such a transparent compound as this, the two morphemes
dix
an
d
neuf
, being words (and hence signs) on their ow
f
f
n
, must surel
y
still count as si
g
ns
(one ma
y
think). But no, sa
y
s Saussure:
dix-neuf
does not contain parts that are si
g
ns
f
any more than
vingt
‘twenty’ does (part 2, chapter 6, section 3; 181; 130). The
t
d
iff
e
r
e
n
ce be
t
wee
n
dix-neuf
an
d
f
v
ing
t
,
as he
p
resents it, involves a new distinction:
between signs that are motivated and
s
igns that are unmotivated. The sign
vingt
is
t
u
nmotivated in that it is purely arbitrary: the sounds (or letters) that make it up give
no clue to its meaning. The sign
dix-neuf
however, contains subunits which give
f
f
c
lues to its meaning that could hardly b
e
stronger. Even so, according to Saussure,
10
A
NDRE
WC
AR
S
TAIR
S
-
M
CC
ARTHY
dix-neuf
is still a single sign on the same plane as
f
vingt
or
t
neuf
or
f
soixante
-
dix
‘seventy’ (literally ‘sixty-ten’). It is the
valeur
of
r
dix-neuf
in the system of French
f
n
umber words that imposes on it the status of a unitary sign, despite its semantic
t
ransparency. (Saussure might also have added that this transparency, real though it
i
s, depends on a convention that belongs to French
l
angu
e
,
not parol
e
:
th
e
co
n
ve
nti
o
n that
co
n
c
at
e
nati
o
n
o
f
dix
an
d
neuf
means ‘ten plus nine’, not ‘ten times
f
n
ine’ or ‘ten to the ninth power’, for example. His ne
g
lect of this point reflects his
g
eneral ne
g
lect of s
y
ntactic and s
y
nta
g
matic convention
.
7
Similarl
y
, the En
g
lish
p
lural for
m
s
hi
p
s i
s
m
o
ti
v
at
ed bec
a
use
it ‘
r
ecall
[
s
]
a whole series like
f
la
g
s, birds,
books
,
etc.’
,
while
men
an
d
s
hee
p
ar
eu
nm
o
ti
v
at
ed bec
a
use
they ‘recall no parallel
c
a
ses
’
.
The
p
lural suffix -
(
e
)
s is, in the English-speaking world, among the first half-
d
ozen ‘morphemes’ that every beginning student of linguistics is introduced to. Yet
for Saussure it does not count as sign; it is merely a reason for classifying the words
t
hat it appears in (
s
hips
,
flags
etc.) as relatively motivated
signs rather than purely
d
arbitrar
y
ones. There is thus a strikin
g
discrepanc
y
between the word-centred
a
pp
roach to com
p
lex words,
p
redominant in the work of the
p
ioneer structuralist
Saussure, and the mor
p
heme-centred a
pp
roach that
(
as we shall see
)p
redominated
among his structuralist successors. In secti
o
n 2 I
w
ill
ou
tlin
e
th
e
attra
c
ti
o
n
s
an
d
p
itfalls of mor
p
heme-centred a
pp
roaches.
2. M
O
RPHEME AND
WO
RD
Saussure recognised some of the difficu
l
ties inherent in using ‘word’ as a
t
echnical term (part 2, chapter 2, section 3). Nevertheless, when illustrating his
n
otion ‘si
g
n’, he chose lin
g
uistic units that in ordinar
y
usa
g
e would be classified as
words
,
such as Latin
a
r
bor
‘tree’ and French
r
juger
‘to judge’ (part 1, chapter 1,
r
s
ection 1;
p
art 2, cha
p
te
r
4
, section 2). This ma
y
be lar
g
el
y
because the lan
g
ua
g
es
from which he drew his examples were nearly all well-studied European languages
with a long written history and a tradition of
grammatical and lexical analysis in
f
t
erms of which the identification of words
(
in some sense
)
was uncontroversial.
However, accompanying the theoretical dev
e
l
opments in linguistics in the early
t
wentieth century was an explosion in fieldwork on non-Indo-European languages,
particularly in the Americas and Africa. In these languages, lacking a European-style
t
radition of
g
rammatical description, identif
y
in
g
words as lin
g
uistic units often
s
eemed
p
roblematic. In fact, there was a stron
g
current of opinion accordin
g
to
which the word deserves no special status in lin
g
uistic description, and in particula
r
n
o special status warranting a distinction
be
t
wee
n th
e
int
e
rnal
s
tr
uc
t
u
r
eo
f
wo
r
ds
(
‘morphology’) and the internal structure of phrases and sentences (‘syntax’). As
Malinowski put it, ‘isolated words are in fact only linguistic figments, the products
o
f an advanced linguistic analysis’ (Malinowski 1935: 11, cited by Robins 1990:
154). So what units are appropriate as tools for a preliminary linguistic analysis? It
see
m
ed
nat
u
ral t
o
an
swe
r
:
th
ose u
nit
s
tha
t
are clearly indivisi
t
ble grammatically and
7
I owe this point to Harris (1987: 132).
B
ASIC TERMINOLOGY
11
l
exically, or, in other words, units of the kind that we provisionally labelled
‘morphemes’ in section 1. Thus, despite Saussure’s leaning towards the word-as-
s
ign position, the experience
o
f fieldwork on languages unfamiliar to most European
and American scholars imposed a prefere
n
c
e for a version of the morpheme-as-si
g
n
position.
W
here, then, does the morpheme-as-si
g
n position leads us? Let us recall first the
S
aussurean norm of what constitutes a
s
ignifian
t
:
a sequentially ordered string of
s
ounds, such as Latin
[
a
r
b
o
r
](
s
p
elled
a
r
bo
r
)
or French
[
<
y
<
e
](
s
p
elled
j
u
g
er
)
, such
t
hat ever
y
unit of parol
e
is anal
y
sable exhaustivel
y
as a strin
g
of
s
i
g
ni
f
iants (part 1,
c
hapter 1, section 3). What we will observe is a temptation towards si
g
ns with
s
ignifiants that deviate progressively further from this norm. The analyses that I will
discuss are based on an approach to morphemes that was expounded in particular by
Z
ellig S. Harris (1942), Charles F. Hockett (1947), Bernard Bloch (1947) and
Eugene A. Nida (1948). None of these
e
xplicitly espouses the morpheme-as-sign
position, because none of them cites Saussur
e
. However, the issues that they discuss
c
an all be seen as prima facie difficulties for that position. The fact that all these
re
f
e
r
e
n
ces
ar
ec
l
us
t
e
r
ed
m
o
r
e
than hal
f
a century ago reflects the replacement of
f
m
orpholo
gy
b
y
s
y
ntax at the centre of
g
rammatical theor
y
-construction.
N
evertheless, I will comment in section 3 on uses of the term ‘morpheme’ since
about 1960.
2
.1 Case study: English noun plural forms (part 1)
For Saussure
,
as we have seen
,
the
-s
su
ffix
of
flags
f
an
d
s
hips is not a sign but an
e
l
e
m
e
nt that r
e
n
de
r
s
th
ose wo
r
ds
r
e
la
t
ively motivated, by contrast with
men
an
d
s
heep
.
Let us sa
y
instead that this
-s
suffix is indeed a si
g
n, with the
signifié
‘plural’.
W
hat is its si
g
ni
f
ian
t
?
So far as En
g
lish spellin
g
is concerned, the answer is simple.
W
hen we turn to phonolo
gy
, however, we encounter our first stumblin
g
-block. In a
c
onventional
p
honemic transcri
p
tion for thes
e
two words, the suffix will a
pp
ear in
t
wo different shapes, /
z
/
an
d/
s
/,
(/
fl
æ
I
z
,
5+
ps
/), and there is
y
et a third shape, eithe
r
/
+
z
/o
r
/
z
/, according to dialect, found in words such a
s
r
oses
,
ho
r
ses
,
chu
r
ches
an
d
judges
.
8
Must we then recognise three different signs with the same
s
ignifié
?S
uch
an analysis would place these three signs
o
n a par with sets of synonyms such as
courge
tt
es an
d
zucchini
,
o
r
n
earl
y
an
d
almost
.
That is hardly satisfactory, because it
neglects the role of phonology in determining the complementary distribution of the
t
hree sha
p
es:
/
+
z
/ a
pp
ears after strident coronal sounds, while elsewhere
/
z
/ a
pp
ears
aft
e
r
vo
i
ced sou
n
ds
an
d/
s
/
aft
e
r
vo
i
ce
l
ess o
n
es.
It was in relation to patterns such as this that the term ‘allomorph’ was first
i
ntroduced in morpholo
gy
. The intended parallel with the notions ‘phoneme’ and
‘allophone’ is evident. Just as sounds that are phoneticall
y
similar and in
8
In m
y
dialect, the third shape is
/
+
z
/
,
so that
taxes
sou
n
ds
th
es
am
e
a
s
taxis
,
but
roses
sou
n
ds d
iff
e
r
e
nt
f
r
om
Rosa’s
.
For many speakers of other dialects, the homophony pattern is the other way round. The
examples that I will discuss fit my own dialect, but
similar examples can easily be constructed to
t
m
ake the same point for speakers with the other homophon
y
pattern.
1
2
A
NDRE
WC
AR
S
TAIR
S
-
M
CC
ARTHY
c
omplementary distribution count as allophones of one phoneme, so individually
m
eaningful units that are not divisible into smaller meaningful units, provided that
t
hey are synonymous and in complementary di
s
tribution, count as allomor
p
hs of one
m
orpheme. And
j
ust as it is the allophones of a phoneme that get pronounced, rathe
r
t
han the phoneme itself, a morpheme is likewise not pronounced directly, but
r
epresented in the speech chain by whichever
o
f its allomorphs is appropriate for the
c
ontext. This applies even to morphemes that have the same shape in all contexts,
because there is no reason in principle wh
y
a morpheme should not have onl
y
one
allomorph,
j
ust as a phoneme ma
y
have onl
y
one allophone.
Notice, however, that that
p
hrase ‘individually meaningful units that are not
d
ivisible into smaller meaningful units’ is lifted from my provisional definition of
‘mor
p
heme’ in section 1. It seems, then, that our ex
p
loration of the mor
p
heme-as-
s
ign position has led us already to a dilemma. If the units
/
+
z
/
,/
z
/
an
d/
s
/
ar
e
Saussurean signs, just like the units /
¡
n
/
(
un-
)
, /
h
e
l
p/
(
h
elp
)
,
/
f
7
ff
l
/
(
-ful
) and /
ll
n
/
/
+
s
/
(
-
ness
) that served to introduce the ‘morpheme’ notion in section 1, then we must
co
n
cede
that th
eu
nit
s
that
de
s
erve ‘sign’ status, as an
a
l
ternative to words
,
are not
after all mor
p
hemes but allomor
p
hs of mor
p
hemes
.
9
Furthermore
,
if
/
+
z
/
,
/
z
/
an
d/
s
/
ar
e
all
s
ignifiants of signs whose
signifié
is ‘plural’, the morpheme that they all
belong to seems somehow superfluous from the point of view of the Saussurean
sign, constituting neither a
signifiant
n
o
r a
t
s
ignifié. On the other hand
,
if we wish to
c
ontinue to say that it is morphemes that are signs, rather than allomorphs, we must
depart from the Saussu
re
an
doc
trin
e
that a
signifiant
is a linearly ordered string
t
within the speech chain (
/
+
z
/, for example), and say instead
that it is, or may be, a set
d
o
f linearl
y
ordered strin
g
s in complementar
y
distribution (
/
+
z
/
,/
z
/
an
d/
s
/
,
in this
i
nstance
).
Th
e
fa
c
t that th
ed
i
s
tri
bu
ti
o
n
o
f th
es
e
allomorphs is phonolo
g
icall
y
conditioned
m
a
y
su
gg
est an escape from this dilemma. If the choice between the three
allomorphs is determined purely by constraints of English phonology, then perhaps
w
e can say that, in phonological terms at least (although not phonetic), we really are
d
ealing with only one string within the speech chain, not three. If so, the
p
roblem of
m
ultiple
s
ignifiants disappears, and the plural
-s
su
ffix
co
nf
o
rm
s
t
o
th
e
n
o
rm f
o
r a
Saussurean sign
.
The stumbling-block is not quite so easily surmounted, however. English
phonolo
g
ical constraints do not suppl
y
a conclusive verdict on which allomorph is
appropriate in all contexts. There are man
y
co
nt
e
xt
sw
h
e
r
e
m
o
r
e
than
o
n
eo
f th
e
t
hree allomorphs is phonologically admissible, and some contexts where all three
are. Consider the noun
pen
/
pen
//
/. Its plural form is /
pe
n
z
//
/, compl
y
in
g
with the
generalisation that the voiced form of the suffix appears after voiced sounds (othe
r
t
han coronal stridents
)
. But this is not b
e
cause the alternative suffix shapes
y
ield bad
phonotactic combinations. Both /
pe
ns
/
/
/
an
d/
¥
pe
n
+
z
/ are phonologically wellformed, and
indeed both exist as words (
pence
(
(
an
d
pennies). So something more than pure
phonotactics is at work in the choice between the three allomorphs. Only in terms of
a
phonological theory more sophisticated than any available in Saussure’s time (fo
r
9
This is the view defended b
y
M
e
Đþ
uk
(
1993-2000
)
.