?tekauer Pavol, Lieber Rochelle. Handbook of Word-Formation
Подождите немного. Документ загружается.
PROD
U
CTIVITY
:
T
HEORIES
321
e
arlier made by Aronoff (1976) that the most productive formation-types are neve
r
l
isted in the mental lexicon, but are processed in real time.
6. NAT
U
RAL M
O
RPH
O
L
OG
Y
Natural Morpholo
gy
1
is a theor
y
of morpholo
gy
in which morpholo
g
ical
behaviour is seen as bein
g
contin
g
ent on co
g
nitive behaviour and
g
eneral semiotic
principles. Natural morpholo
g
ical phenomena are those which are common cross-
l
inguistically, which resist erosion either diachronically or in aphasias, and which
are easily acquired both by children learning their first language and in the
development of creoles. Various general semiotic principles govern morphological
s
ystems, among them the principle of
constructional iconicity
(more meaning is
y
m
ost naturally reflected in mo
r
e form) and principles of
t
ransparenc
y
(the clarity of
t
he semantic and phonological identity of fo
r
ms, bases and affixes); generally the
m
ore constructionall
y
iconic a word is, the more natural it is and the more
t
ransparent the morpholo
gy
, the more natural the word is. Thus
s
hee
p
as a
p
lural
form is not constructionall
y
iconic (compare
ewe
·
s
where the extra meanin
g
of
‘
plural’ is reflected by an extra morph), and
dividedness
is more trans
p
arent than
division
because the phonological integrity of
divide
i
s
maintain
ed
in
dividedness
,
bu
t n
o
t in
division
.
The basic approach to productivity within Natural Morphology is that it is a by-
product of naturalness (Mayerthaler 1981: 124-140). If this is true we would expect
word-formation b
y
internal modification to be less productive in En
g
lish than word-
formation b
y
affixation (since internal modification which links
sing
t
o
g
song
,
rise
t
o
raise
an
d
believe
t
o
belief
) is less constructionally iconic than affixation is. This is
f
f
c
ertainly true. The cited examples of word-formation by internal modification in
English are rare and synchronically unproductive. Even the status of stress-shift
(
which links the noun
‘import
to the verb
t
i
m
’p
or
t
) is unclear in current English. It
also suggests that processes of word-formation which cause less morphophonemic
v
ariation for the base should be more p
r
oduc
ti
ve
than th
ose w
hi
c
h
c
a
use
m
o
r
e.
Thi
s
i
s probably also true, though less obviously so. Smith (1912: 60), for instance,
c
omments that ‘b
y
far the most active of
o
ur affixes are Greek in ori
g
in.’ These
i
n
c
l
ude
-
ic
,
-
ist
and
t
-
i
z
e
,
which cause velar softenin
g
in the base, so that
i
tali
[
k
]
beco
m
es
i
tali
[
s
]
iz
e
,
for example. While more recent tallies su
gg
est that these affixes
are less productive than Smith suggests,
t
hey are nevertheless still productive in
c
ontemporary English, despite having competitors such as -
ish
,
-
er
and conversion
r
r
espectively which do not have these effects.
C
ertainly these effects are not obvious if they are considered in particula
r
domains. For example, -it
y
,
which changes the stress of any base to which it is
attached
,
and thus reduces the transparency (and so the naturalness) of any formation
i
n which it appears, is virtuall
y
universal on bases endin
g
in -
able
(
profitability
(
(
i
s
m
uch more likel
y
that pro
f
itableness). If we simpl
y
consider the domain of
-able
-
1
Cf. W. Dressler’s cha
p
te
r
i
n thi
svo
l
u
m
e
f
o
r th
ede
tail
so
f th
e
t
r
e
atment of WF within the framework of
Natural Morpholo
gy
.
3
22
L
A
U
RIE BA
U
ER
adjectives, therefore, -it
y
will appear to be productive beyond its naturalness,
whereas if we consider the language as a whole, -
ness
,
which is more natural than
-it
y
,
is also more
p
roductive.
The structure of the individual language can also play a role in what counts as
n
atural for that language. In the case of English derivation, we can see that class-
c
hanging morphology is overwhelmingly suffixal (or, in another terminology,
En
g
lish morpholo
gy
is ri
g
ht-headed) and so the use of prefixes to create denominal
v
erbs is unnatural in the s
y
stem of En
g
lish and correspondin
g
l
y
restricted in its use.
Examples such as
deth
r
one
,
disth
r
one
,
enth
r
one
an
d
unth
r
one
ha
ve
all
bee
n
used
at
d
ifferent periods of English (some of them still persisting), but the pattern in
enth
r
one
is now no longer available, the pattern shown in hos
p
italize being used
i
nstead. Something which appears to be developing in the wrong direction is the
apparently left-headed patt
e
r
n
o
f m
od
ifi
e
r
ssuc
h a
s
o
estrogen-only (pill), fruit-only
(jam
)
, which are frequent – at least in jour
nalistic texts – though apparently never
rr
i
nstitutionalized enough to make it into di
c
tionaries. These go against the very
g
eneral principle for En
g
lish compounds to be ri
g
ht-headed (Bauer & Renouf 2001).
As well as productivity being influenced
by transparency and constructional
d
i
conicit
y
, Dressler and Ladán
y
i (2000) su
gg
est that base-t
y
pe pla
y
s a role. Fo
r
example, a base which is a new loan is more likely to be used productively in a
c
oinage than a base whic
h
i
s
an a
bb
r
ev
iati
o
n
.
7
. KIPARSKY
(
1982
)
Kiparsky (1982) presents what was then the new field of
l
exical phonolog
y
in
s
ome detail. In the course of doing this, since lexical phonology is concerned with
the interpla
y
between morpholo
gy
and phonolo
gy
, he inevitabl
y
discusses
m
orpholo
gy
to a lar
g
e extent. It is here assumed that the reader has a basic
familiarit
y
with lexical phonolo
gy
.
B
ecause of the way in which a
l
evel-ordered morpholog
y
works
,
individual
words cannot be blocked by words created with affixes on a later level: that is, a
word containing a level 2 affix such as -
e
r
c
annot block a level 1 process such as
ve
r
b
ĺ
noun conversion; on the other hand, the
p
resence of
cook
N
k
k
c
an
b
l
oc
k
cooke
r
w
ith the same meaning (note that
cooker
as an established word is an instrument – or
r
an apple – not a person). This is part of the Elsewhere Condition on rule-ordering.
Kiparsky (1982: 8) explains that
From this it follows in turn that amon
g
processes in a blockin
g
relationship, those with
r
estricted applicabilit
y
have to be ordered before those with
g
eneral applicabilit
y
. This
explains wh
y
processes at later levels are also t
y
picall
y
more productive than
functionally related processes at earlier levels.
Part of this productivity is because idiosyncrasy tends to d
e
v
elop at the lower-
n
umbered levels in a level-ordered system. This does not preclude idiosyncrasy at
l
evel 2 or above
(
for exam
p
le,
h
i
g
hness is not s
y
non
y
mous with
height
but is a term
t
o
f address
)
, but means that forms created b
y
processes at hi
g
her levels are more
l
ikel
y
to be easil
y
interpretable. This draws a connection between semantic
PROD
U
CTIVITY
:
THEORIES
323
t
ransparency and productivity which has already been seen in Natural Morphology,
but attempts to explain it by the form of the model. One of the corollaries of this
l
ack of idiosyncrasy at higher levels is that once affixes have been added,
s
ubsequent affixation tends to be more re
g
ular than is affi
x
ation to simple bases. We
c
annot predict in an
y
lin
g
uistic wa
y
that we should have exposure, imposition,
proposal
rather than some other mixture of base and suffix, but once we have
l
-
i
z
e
c
reating the verb, we can predict what the nominalization of that verb will be.
8. VAN MARLE
(
1985
)
V
an Marle’s theory of productivity is fi
r
mly based in the Schultink tradition of
Dutch scholarship. However, Van Marle is not content with the one-affix-at-a-time
approach that typifies the majority of scholarship on word-formation. He sees
affix
es
an
do
th
e
r
wo
r
d
-f
o
rma
t
ion processes as being in competition with each othe
r
i
n wa
y
s determined b
y
the structure of the derivational paradi
g
ms in which the
y
o
ccur. For Van Marle, morpholo
g
ical productivit
y
is a subset of
m
orpholo
g
ical
creativit
y
,
creativit
y
includin
g
much that is be
y
ond the normal rules o
r
m
orphological structure. Int
e
r
estingly, Van Marle include
s
affix
subs
tit
u
ti
o
n in thi
s
c
lass. Affix-substitution is arguably rather more common in Dutch than in English.
I
n English we can illustrate with forms like those in Table 1 where the relationship
s
eems to arise through the substitution of on
e
affix for another rather than by the
c
ompositional application of affix-meaning to base-meaning.
c
apitalism capitalist capitalistic
hu
mani
s
m h
u
mani
s
t h
u
mani
s
ti
c
nati
o
nali
s
m nati
o
nali
s
t nati
o
nali
s
ti
c
p
erfectionism
p
erfectionist
p
erfectionistic
p
luralism
p
luralist
p
luralistic
p
ositivism
p
ositivist
p
ositivistic
s
a
d
i
s
m
s
a
d
i
s
t
s
a
d
i
s
ti
c
Ta
b
l
e1
English words created by suffix-substitution
C
ontrast the examples in (1) with
i
mpressionisti
c
,
which in current usage does
no
t r
efe
r t
o
i
mpressionism
,
(the adjective in this sense is
i
mpressionis
t
)
, and Baptis
t
which is not directl
y
related to
b
a
p
tism in the same kind of wa
y
(
b
a
p
tism an
d
b
a
p
tize
are probabl
y
related b
y
affix-substitution, as are
m
a
g
netism an
d
ma
g
netiz
e
)
. The
i
mplication for Van Marle is that such fo
r
m
s
ar
e
t
obee
x
c
l
uded
fr
o
m th
e
fi
e
l
do
f
productive relationships by d
e
finition. They show the importance of paradigmatic
s
tructure, and they show
m
orphological creativity, but
n
ot, by definition for Van
Marle, productivity.
Equally, as far as Van Marle (1985: 59) is concerned, any formation which
i
nvolves non-native morphology is intentionally formed, and thus not productive
m
orpholo
gy
. This too is controversial, perhaps even more so in En
g
lish than in
3
2
4
LA
U
RIE BA
U
ER
Dutch
(
where native and non-native affixes tend
to be
rath
e
r m
o
r
e
r
es
tri
c
t
ed
t
ob
a
ses
o
f equivalent status than is the case in English). If we consider patterns of word-
formation with non-native affixes in Englis
h(
non-affixal methods of formation can
probably all be seen as native), we have to include prefixes such as
ex-
,
in-
(negative),
inte
r
-
,
mal-
,
pos
t-
,
pre-
,
r
e
- an
dsu
ffix
es suc
h a
s
-
al
]
A
,
-
al
]
N
,
-
an
,
-
a
r
,
-
ation
(
and
-
ition
,
-ution
)
, -esqu
e
,
-
ess
,
-
ette
,
-
ic
,
-if
y
,
-
ism
,
-
ist
,
-
i
z
e
,
-it
y
,
-
ment
,
-
ous
and so on. While there is no doubt some si
g
nificance in the fact that so man
y
of
these suffixes are vowel-initial
(
Raffelsiefe
n
1999), the conclusion that the
y
are all
unproductive because the
y
can onl
y
be u
s
ed deliberatel
y
is not onl
y
counter-
i
ntuitive, but seems to be contrary to the findings of scholars like Aronoff and
Schvaneveldt (1978), Anshen and Aronoff (1981), Baayen and Lieber (1981). At the
v
ery least, we seem to be dealing with a different underlying concept of what
productivity is all about if w
e
accept Van Marle’s position here.
9
. CORBIN
(
1987
)
C
orbin (1987) is not primarily concerned with productivity as a phenomenon, but
m
akes a ver
y
important distinction and provides a useful terminolo
gy
: that between
disponibilité
(translated by Carstairs-Mc
C
arth
y
1992 as ‘availabilit
y
’) and
rentabilité
(translated as ‘profitability
’
). A morpholo
g
ical process is
d
isponibl
e
if th
e
s
peaker is able to use it in the construction of a neologism, filling a lexical gap
(
Corbin 1987: 177
)
. It is
rentable
if it is actually used to produce a large number of
s
uch words. Thus on the evidence of Barnhart, Steinmetz & Barnhart
(
1990
)
, which
l
i
s
t
s
awhir
and
r
aclutte
r, amongst others, as neologisms, we might say that the non-
n
egative prefix
a-
is available in English (or was in the period they cover).
N
evertheless, the prefix
a-
has never been profitable, and the number of words
c
reated b
y
it remains small.
A
lthou
g
h this distinction, and the terminolo
gy
that
g
oes alon
g
with it, is credited
h
ere to Corbin, what I interpret as fundamentally the same distinction was
i
ndependently drawn by Kastovsky (1986). Kastovsky’s distinction, though, is
phrased as having all matters that can be seen as matters of competence determining
what Kastovsky calls ‘rule scope’ and, opposed to that ‘application rate’ which is a
performance factor measured in terms of number of outputs from a particula
r
pattern.
10
.
BAAYEN
B
aa
y
en is more concerned with the development of a measure o
f
productivit
y
than he is with theorizin
g
the notion. Nevertheless, an
y
measure has behind it an
i
mplicit theory of what it is that is bei
n
g measured, and Baayen also makes reference
to earlier studies of productivity. He points out (1992: 109-10), for example, that the
productivity of various mo
r
phological processes may
d
iffer according to socio-
pragmatic environment such as register, or
whether the recorded language is written
r
PROD
U
CTIVITY
:
T
HEORIES
325
o
r spoken, as well as according to linguistic
re
a
so
n
ssuc
h a
s
th
e
transparency of the
process involved.
B
aayen attempts to measure the productivity of morphological processes on the
basis of a lar
g
e parsed
corpus
of natural lan
g
ua
g
e. In Baa
y
en and Renouf (1996: 69),
i
t is ar
g
ued that ‘[d]ictionaries, unfortunatel
y
, are not a reliable source for stud
y
in
g
m
orpholo
g
ical productivit
y
’ because of the co
m
mercial and practical aims of real
d
i
c
ti
o
nari
es
an
dbec
a
use o
f th
e
fa
c
t n
o
t
ed
by Aronoff (1976) that
d
dictionaries do not
t
n
ote words formed by the most productive morphological processes. Baayen and
Renouf comment specifically on the suffix
-
l
y
]
AD
V
,
which shows few new forms in
dictionaries of neologisms, but which is used widely in a corpus.
B
ecause even a ver
y
lar
g
e corpus of t
e
x
ts (Baa
y
en is t
y
picall
y
operatin
g
with a
c
orpus of around 20 million runnin
g
words)
c
annot have ever
y
sin
g
le word eve
r
u
sed with ever
y
affix in it, the corpus has to be seen as a sample of the lan
g
ua
g
e as a
whole. Baayen argues that a guide to the number of new forms we would expect to
find in the language as a whole is the number of relevant hapax legomena (usually
j
ust called ‘hapaxes’, that
i
s words which occur only o
n
c
e in the corpus) which are
found in the sample. If there are very few words formed with a particular affix, fo
r
e
xample, and it is never used to produce new words, then we might expect to find
m
ost of those which do exist in a large corpus. But if the affix under consideration
c
an be used freel
y
to make new words, we would expect to find man
y
different
words in our sample, and we would expect to find more of them occurrin
g
onl
y
o
nce. In this view
,
it does not matter if
t
he words which occur onl
y
once are actuall
y
familiar words or not: they represent the rate at which new words are being coined in
t
he language as a whole. Baayen thus creates measures which centre round the
number of hapaxes in a corpus. His notion of productivity is rather similar to that put
forward by Aronoff, but operationalized so that we count hapaxes rather than
possible words and types in the corpus rather than actual words. Although Baayen
provides a number of different formulae for the measurement of productivit
y
in his
work (e.
g
. Baa
y
en 1989, 1992, 1993), the one which has most frequentl
y
been
discussed in subse
q
uent work is the one
g
iven in (1992), namel
y
P
=
n
1
N
w
h
e
r
e
P
is the productivity of a particular morphological process in a given corpus,
n
1
is the number of hapaxes formed b
y
that process in the same corpus, and
N
is the
N
t
otal number of tokens formed in the same corpus b
y
the same process. Note that
while Baa
y
en’s measurement allows comparative measures of productivit
y
to be
g
iven for two morphological processes in the same corpus, it does not allow
different cor
p
ora to be com
p
ared
(
and thus, for instance, does not allow com
p
arisons
o
f the productivity of English -
e
r
,
German -
er
and Dutch
r
-
e
r
)
.
An ad
j
ustment can then be made to take absolute frequencies into account, since
part of the reason we think some morphological operations more productive than
o
thers is that they produce a great many words. For instance, the prefix
st
ep-
(
as in
step-mother) does not produce ver
y
man
y
words, and is probabl
y
not considered to
be particularl
y
productive, while the a
g
entive -
er
(as in
r
mu
r
de
r
e
r) is thou
g
ht to be
3
2
6
LA
U
RIE BA
U
ER
m
uch more productive partly because we
m
ee
t m
o
r
ewo
r
ds co
ntai
n
ing the suffix. It
should be noted,
p
arenth
e
t
ically, that we do not necessarily know whethe
r
s
te
p-
,
if
t
here are no more possible bases it could ever apply to, is 100% productive o
r
u
nproductive. If we conclude that it is 100% productive over the available bases, this
i
s still not very important as far as Eng
l
i
s
h i
sco
n
ce
rn
ed bec
a
use o
f th
e
l
ow
n
u
m
ber
o
f
b
a
ses
in
vo
l
ved.
The formula which Baa
y
en (1993: 193) proposes as providin
g
a better wa
y
o
f
r
ankin
g
processes in terms of their comparative productivit
y
is the measure he labels
P
*, called ‘the hapax-conditioned de
g
ree o
f
productivit
y
’. This measure is
g
iven in
t
he following formula:
P
*=
n
1,
E
,
t
h
t
In this formula
,
n
1
is still the number of hapaxes formed by the relevant process
attested in the cor
p
us,
E
indicates the relevant morphological category and
E
t
i
ndicates the number of tokens in the cor
p
us, while
h
t
is the total number of hapaxes
t
o
f all types in the corpus. This formula ca
n
thus be read as ‘the hapax-conditioned
degree of productivity of a given process in a corpus can be defined as the numbe
r
o
f words of the appropriate morphological category appearing
j
ust once in the
c
orpus divided b
y
the total number of hapaxes appearin
g
in the corpus’. In othe
r
words, this measure asks ‘What proportion of hapaxes in the corpus are formed
u
sin
g
this particular morpholo
g
ical process?’, while the first measure asks ‘What
proportion of words formed by this particular process are hapaxes?’. In both cases,
t
he hapaxes are an indirect representation of the use of the morphological process in
t
he coining of new words. Here we might say we are viewing two different aspects
o
f profitability, two which need not coincide. The measure
P
co
n
s
i
de
r
s
h
ow
frequentl
y
the words we meet in a particular morpholo
g
ical cate
g
or
y
are new words;
t
h
e
m
e
a
su
r
e
P
*
considers what proportion of new words are created by the process
i
n which we are interested. This suggests that Corbin’s ‘profit
a
bility’ might need to
be
f
u
rth
e
r
subd
i
v
i
ded.
11. PLAG
(
1999
)
There are two aspects to Pla
g
’s (1999) discussion of productivit
y
which are of
particular interest. The first of
t
hese is that Pla
g
ar
g
ues that a
d
ictionar
y
search
,
or
,
m
ore specifically, a di
c
t
ionary search using
T
he Oxford English Dictionar
y
,
can
provide solid evidence on the productivity of particular morphological processes in
different periods of history. In this, he a
r
g
ues directly counter to the view taken by
B
aayen (e.g. in Baayen a
n
d Renouf 1996: 69). Secondly, for Plag, implicitly at any
r
ate, productivity is related to
domains
,
but to rather larger domains than those
e
nvisaged by Aronoff (1976).
In re
g
ard to the use of the
O
E
D
as a database, Pla
g
(1999: 96-100) ar
g
ues that
u
nlike smaller dictionaries
,
the
O
E
D
does make some attempt at universal covera
g
e
(thou
g
h he admits that words in -
ness
appear to have been missed b
y
readers for the
PROD
U
CTIVITY
:
T
HEORIES
32
7
O
E
D
). He also argues that because the
O
ED gives dates of first citation, we can see
the periods during which a particular morphological process has been productive,
and we can look away from the many words that are or have been used in the
l
an
g
ua
g
e which are lexicalized and do not represent productive processes. And
finall
y
, while a
g
reein
g
that the database on which the
O
E
D
is founded is ver
y
v
a
g
uel
y
defined, he points out that it is man
y
times lar
g
er than the kind of corpus
t
hat is employed in corpus-based studies.
Although Plag does consider the morphologi
c
al processes he covers individually,
at a more general level he is concerned with all processes which give rise to derived
v
erbs in modern English. Thus he is able to look at the way in which this wide
domain is divided up among the available morphological processes. To some extent,
t
h
e
r
esee
m
s
t
obe
a
se
manti
c
fa
c
t
o
r at work here. He argues, for instance, that -
ate
when it is added as a verb-creatin
g
affix, is used productivel
y
onl
y
in words
denoting chemical processes (
fluorinate/fluoridate, metalate
(
(
), which distin
g
uishes
t
hi
s
affix fr
om
-i
fy
an
d
-
i
z
e
. These latter two suffixes, however, are s
y
non
y
mous, and
which is used is determined by the phonological structure of the complex word in
which they might occur. Plag uses
Optimality Theory
to prioritize the various factors
y
which influence the form of the finished words, and shows that basically -if
y
an
d
-
i
z
e
are in complementary distribution
.
Where they are not (e.g. in
d
andyize
or
d
andif
y
)
i
t is because the phonological requirement
s
do not distinguish between the two
possible outputs, and both are found. So for Pla
g
the
r
elative productivit
y
o
f th
ese
two is determined b
y
the phonolo
g
ical requirements on En
g
lish words and the
phonolo
g
ical structures of the words whic
h
are potentiall
y
bases for each of the
processes. The productivity of
-ate
,
in contrast, is determined by the need for a
particular type of technical derivative.
Relative productivity, in this view, is the result of heterogeneous causes. It can
be measured in any particular period by the numbers of relevant forms listed by a
dictionary such as the
O
E
D
.S
ince the
O
E
D
has probably to be seen as a unique
l
exicographical resource, it is not clear how far such measures can be used in othe
r
l
an
g
ua
g
es (thou
g
h Bolozk
y
1999 ar
g
ues that dictionar
y
measures are useful even in
l
an
g
ua
g
es with a less rich lexico
g
raphic tradition).
12. HAY
(
2000
)
Some of the findings of Hay (2000) hav
e
been published in more accessible for
m
as Hay (2001). Others are still in the pipeline.
Hay attributes productivity to the interaction of two factors:
r
elative frequenc
y
of
de
ri
ved wo
r
d
an
db
a
se
an
d
phonotactic patterns at morph boundaries. Each of these
w
ill
be de
alt
w
ith in t
u
rn
.
According to Hay, what is important in frequency terms is not the absolute
frequenc
y
of individual affixes or words, but the relative frequenc
y
of the base and
t
h
ede
ri
v
ati
ve.
Th
e
r
e
ar
e
instances where the base is more fre
q
uent than the
derivative
(
taste
is a far more commonl
y
used word than
tasteless
,
for exam
p
le
)
and
i
n
s
tan
ces w
h
e
r
e
th
ede
ri
ved wo
r
d
i
s more common than its base
(
exactl
y
i
s
far m
o
r
e
co
mm
o
n that
exact
). Where we find affixes whose base is regularly more common
t
3
2
8
LA
U
RIE BA
U
ER
t
han the affixed form, we are likely to ha
v
e
p
roductive affixes, and where we have
affixes which regularly produce words which are commoner than their bases, the
affix is likely to be non-productive, Hay claims.
This feature also correlates with the second feature. Hay considers the
phonotactics which arise over morph boundaries. A consonant sequence like /
nh
//
/
does not occur within a morpheme, and thus the sequence
/
n
h
/
,
when it occurs in
wo
r
ds
lik
e
inhuman
,
is a major sign that there is a morphemic boundary between the
/
n
/
an
d
th
e/
h
/. On the other hand /
mp
/
/ occurs frequently inside morphemes, as in
l
am
p
, am
p
er
e
etc
,
so that the
/
mp
/w
hi
c
h
occu
r
s
in
i
mprope
r
is not a good signal that
r
there is a morpheme boundary at that point. Hay claims that affixes which tend to
c
ause boundaries which are phonotacticall
y
si
g
nalled, are more productive than
affixes which in
g
eneral do not si
g
nal a morpholo
g
ical boundar
y
in this wa
y
. She
c
laims that this is related to the facts which have been discussed under the headin
g
o
f Level Ordering. Consonant-initial suffixes are more likely to provide phonotactic
boundary signals than vowel-initial ones, and overwhelmingly consonant-initial
s
uffixes make u
p
the suffixes at Level 2
(
the more
p
roductive level
)
in a level-
o
rdered description of English.
W
hat Hay does here, and it is the great strength of her approach, is that she
provides a perceptual basis for
j
udgements about how decomposable or analyzable
words
(
and a
f
ortiori the affixes which are used to make up those words) are. Ha
y
is
also concerned to link this anal
y
zabilit
y
to thin
g
s like the order in which affixes
appear in words. The ps
y
cholin
g
uistic foundation of Ha
y
’s theor
y
makes it
particularly appealing to many linguists. The disadvantage of the theory is that it
fails to account for all of the facts. One of the repeated findings of people who loo
k
at productivity diachronically (e.g. Anderson 2000; Bauer, 2001, forthcoming;
C
owie, 1999) is that productivity of the same affix can change over time. The
phonotactics probably remain the same over fairly long periods of time. All that can
c
han
g
e is the relative frequenc
y
of the deri
v
atives and their bases. Ha
y
would have
to demonstrate that such chan
g
es over time correlate with the productivit
y
chan
g
es.
W
hile this would be a ver
y
useful step, it would seem to downpla
y
the phonotactic
p
rinci
p
le which seems so a
pp
ro
p
riate in the model as it stands.
13. BAUER
(
2001
)
B
auer (2001) does not present a thorough theory of productivity. Rather there is
a discussion of the various factors which go towards productivity and have
so
m
e
tim
es bee
n
co
nf
used w
ith it
.
B
auer presents a careful analysis where he shows that productivity cannot be
s
een as directly equivalent to frequency, transparency, regularity, naturalness o
r
default, although each of these notions may be implicated in productivity.
Productivit
y
cannot be equivalent to
f
requenc
y
because there are
p
rocesses which
c
ontinue to produce new words but at a ver
y
low rate (Bauer 2001: 48 cites
f
o
rmati
o
n
ssuc
h a
s
aclutter
and
r
awhi
r
)
, while, on the other hand, there are
p
rocesses
PROD
U
CTIVITY
:
T
HEORIES
329
which can be attested frequently in a dicti
o
nary but which do not appear to be used
i
n the coining of new forms (Bauer cites -
ment
).
t
T
hat productivity and
t
ransparency
c
annot be equated is
i
llustrated by Aronoff
(
1983
)
in his discussion of nouns i
n
-ibilit
y
(see above), and a
g
ain b
y
-
ment
,
which
h
as become non-productive despite remainin
g
transparent. We can also consider the
c
ase of semantic transparenc
y
.
Lack of semantic transpa
r
enc
y
in an individual word
i
s often taken as a signal of lexicalization of that work. This seems to imply that all
p
roductive word-formation is semantic
a
l
ly transparent and compositional (as
A
ronoff 1976: 45 remarks, ‘productivity goes hand in hand with semantic
c
oherence’). The problem is that much non-productive morphology is also
s
emantically coherent. Again -
ment
springs to mind. A form such as
t
derailment
may
t
n
ot be creatable in the present stage of the language system, but its meaning is
perfectl
y
clear. It is even ar
g
uable that not all productive morpholo
gy
need be
s
emanticall
y
coherent. Without a context, it ma
y
not be clear whether a
choker
(in
r
t
he sense in which it listed as a new word b
y
Knowles 1997) is a person, instrument
o
r location. Only in context does it become clear that it is an instrument used to
c
hoke
(
or slow down
)
traffic in suburb
a
n neighbourhoods. Thus we can argue that
-
er
suffixation, productive though it clearly is, is not very semantically coherent.
r
Regularity
i
s difficult to distinguish from p
r
o
ductivity, if only because the ter
m
i
s used differently by different authors, and it is not clear what it really means. For
s
ome authors, thou
g
h, somethin
g
is irre
g
ular if it is not the ma
j
orit
y
pattern. Yet
m
inorit
y
patterns are not necessaril
y
unproductive, as is shown b
y
Dutch plurals in
-
s,
which remain freel
y
coinable (for i
n
stance the plurals of newl
y
coined
d
imin
u
ti
ves
tak
e
-
s
)
, even if the -
en
marker is the one found in the majority of cases.
Some authors appear to use the term ‘regular’ to mean ‘productive’, and then, of
c
ourse, distinctions can no longer be made.
Similarly, although there is considerable argumentation in the literature that the
default plural on German nouns is -
s
(since this occurs in places where the noun
s
ystem makes no predictions, such as on acronyms like
U
F
O
s
,
phonologically
s
trange formations like A
utos
– strange in that this word ends in an unstressed
/
o
/–
and prepositions or conjunctions used as nouns like A
be
r
s
‘buts’ – compare English
‘but me no buts’), yet this is not the only plural suffix in German which can be used
for new nouns (new feminine nouns regularly take
-
(
e
)
n
)
. A default must be
productive, but not all productive morphological processes must be defaults.
N
aturalness is partly a matter of transparency and other factors mentioned here,
and to the extent that they do not equate with productivity, naturalness does not
equate with it either. The -ibilit
y
example certainly shows that processes which are
n
ot maximally natural can nevertheless be productive.
B
auer also distin
g
uishes between ‘productivit
y
’ and ‘creativit
y
’. The distinction
i
s not original, though the precise definiti
o
n an
d
t
e
rmin
o
l
o
gy may be. Although
B
auer’s analysis differs from the earlier wo
r
k in many ways, the debt to Aronoff
(
1976) and Van Marle (1985) is clear. Although ‘productivity’ and ‘creativity’ are
s
ynonyms for some authors, we can take the
c
r
eative
in creativit
y
literally, and use
t
his term to refer to the less automatic creations, those which are clearly deliberate
and independent of the s
y
stem. These ma
y
be the words of poetr
y
and headline-ese
330
LA
U
RIE BA
U
ER
r
eferred to above, or they may be cases which have the power to begin new
paradigms, such as the reanalysis of
alcoholic
f
ou
n
d
in f
o
rmati
o
n
s
lik
e
chocoholic
.
This leaves the term ‘productivity’ for use with those formations which are clearly
part of the system, namely those parts of word-formation which are rule-governed.
This is in the spirit of Schultink’s (1961) definition of producti
v
i
ty, but avoids the
o
b
j
ectionable reference to inadvertent co
i
nage. Bauer (2001: 66-71) argues against
Schultink (1961) and Van Marle (1985), that conscious formation, includin
g
formations involvin
g
forei
g
n word-formation elements, are to be excluded from the
domain of productivel
y
formed words. Bauer ar
g
ues that ‘conscious’ in this
formulation cannot be practically interpreted, and that if foreign elements are
excluded on the grounds that recognising them involves a certain metalinguistic
facility, the same metalinguistic facility must be employed in the recognition of
n
ative (non-foreign) elements, and the same reasoning should exclude both. Since
the conclusion is absurd, we have to assume that foreign elements in word-formation
c
an indeed be used productively.
This distinction between productivit
y
and creativit
y
means that we need to know
when attestation of new forms (usuall
y
taken as the
s
ine qua non of productivit
y
)
m
a
y
not indicate productivit
y
at all. Bauer (2001: 57-8) provides the followin
g
list:
•
words which occur onl
y
in poetr
y
or hi
g
hl
y
literar
y
texts do not necessaril
y
i
ndicate the productivit
y
of their elements.
•
words which occur only in newspaper headlines do not necessarily indicate the
productivity of their elements.
•
playful formations, where the meaning of a morpheme is apparently disregarded
i
n the creation of a new word, do not necessarily indicate the productivity of
t
h
e
ir
e
l
e
m
e
nt
s.
•
words which occur in the production of a single individual do not necessarily
i
ndicate the productivity of their elements.
•
n
ew technical terms do not necessarily indicate the productivity of the elements
used
t
oco
n
s
tr
uc
t th
e
m
.
•
a single new word apparently showing the use of a particular morpheme is not
s
ufficient to guarantee that the morpheme is productive.
14.
SO
ME THREAD
S
Runnin
g
throu
g
h this whole discussion of productivit
y
are various threads which
we can now attempt to disentan
g
le.
One of these is whether productivity is a gradable/scalar phenomenon or not. Fo
r
Schultink (1961) it is not: either you have t
h
e
ability to create an infinite number of
words according to some pattern or you do
n
ot. Yet for many other scholars there is
at least a gradable aspect to productivity (see, among many others, Bauer 1992).
C
orbin (1987) brings this distinction to the fore, provides a terminological
distinction, and settles the matter by pointing out that there are two distinct questions
i
nvolved here. Althou
g
h Corbin’s terminolo
gy
is not widel
y
used at the moment, it
c
an be applied to subsequent discussion. For example, most of the work of Baa
y
en is