Lingua 18 (1967) 1-34, © North-Holland Publishing Co.

, Amsterdam
No t to be reproduced by photoprint or microfilm without written permission from the publisher

A D J E C T I V E S IN E N G L I S H : A T T R I B U T I O N
AND PREDICATION*)
DWIGHT BOLINGER

SUMMARY

The traditional relative-clause transformation fails to account for many if

not most instances of attributive adjectives. Its attractiveness has possibly
been due to two factors: to our t e n d e n c y to think up easy examples, which
are a p t to have semantically barren words like man whose reference systems
are close to their gTammatical categories, and to the suggestive power of
classical s y l l o ~ m s . One pr~'~f of *~"'-
,~.ur~- is t h a t instead of clearing up ambi-
g u i t y the transformation creates it. There is a clear functional difference
between predicative modification and attributive modification. Two solutions
are offered to account for the restrictions. The first is t h a t be predications
in so far as t h e y are involved at all are of the aspectual t y p e bentemp which
selects adjectives whose meanings can have a temporal spread. The second is
t h a t t w o t y p e s of generation be recognized, one, termed reference-modi-
fication, being in the kernel and allowing for a 'kind of' slot among the
determiners, t h e other, termed referent-modification, being b y w a y of a
predication which is joined b y conjunction rather than b y subordination.

Is the relationship between predicative and attributive adjectives

in English close enough, quantitatively and qualitatively, to
justify describing one transformationally in ~erms of the other?
Assuming that it is, but less so than, say, the relationship between
active and pas.~ive, does it follow that some transformations are
better than others? Assuming that it is not, axe other transfor-
mational descriptions possible, or should attribution be placed in
the grammatical kernel? Or is it better to use one approach for
some attributive constructions and the other for the rest ?
Questions like these need to be discussed informally before
deciding to move in one direction or the other. For all its attractive-
ness, generativists have been a bit rash in adopting without much
*) Fred W. Householder, Jr. and R a n d o l p h Quirk criticized an earlier
draft of this artmle, I am grateful for their help.
2 DWIGHT BOLINGER

debate tile Port Royal description of attributive adjectives, which

can be expressed transforrnationaUy as follows:
I bought the table t
The table was big j ~ I bought the table t h a t was big
I bought the table big ~ I bought the big table
Superficially this looks good; but an example that resembles it
somewhat and that might appear equally attractive is the following:
The poor m a n contributed generously ~ The poor man
generously contributed
Yet the second sentence is more properly based on the same kind
of embedding that produces The poor man's contributing was
generous and It was generous o/the poor man to contribute, i.e.,
The poor m a n contributed [ ~ The poor man generously
it was generous } contributed
since when moved to the left the adverb modifies the sentence as a
whole. The obvious transformation is not always the right one. I
shall t~, to show that this is true of adjectives as well - that most
predicatives with be are fundamentally different from attributives
and that this is reflected in their acceptability orderings, restrictions,
and in other ways.
Before attempting to analyze the difference let us reassure
ourselves that it exists, and not in any marginal sense.

I. SHORTCOMINGS OF be PREDICATIONS
There are m a n y attributive adjectives that are never predicative.
Some ex tmples, without regard to subclasses:
the main reason; *The reason is main
a crack salesman; *]'he salesman is crack
a fond old man; *The old man is fond
a runaway horse; *The horse is runaway
a total stranger; *The stranger is total
Others allow particular attributive uses that lack a predicative
counterpart:
an ~ngry storm; *The storm is angry
a medical man; *The man is medical
 ADJECTIVES IN E N G L I S H 3

There are also adjectives, but fewer of them, t h a t are predicative

but seldom or never attributive, at least in the same sense:
The m a n is asleep; *an asleep man
The m a n is flush ['has a lot of m o n e y temporarily']; *a flush
marl
The girl is sorry; *a sorry girl
(Sorry in its other sense allows attribution but not predication: a
sorry sight; *The sight is serrry). By itself the fact t h a t m a n y more
adjectives are restricted to attributive position t h a n to predicative
position is suspicious; if a n y t h i n g the reverse should be true if we
want to base a t t r i b u t i o n on predication.
Our suspicions are h.loh~e, ~'~ f no d b y another fact, related to one of
the ehieJ[ advantages of generative g r a m m a r : t h a t it enables one to
disambiguatc constructional ambiguity. Instead of clearivg up
ambiguity the traditional predicative-attributive transfo__rmation
creates it. W i t h most adjectives the contrast is subtle, but if we
look first at the perfect participle we can see it in bolder form: The
jewels are stolen is ambiguous as between action (pa ssive voice) and
characteristic (I shall t r y to justify this term lz ter). The stolen
jewels and the jaoels stolen are unambiguous - char~ .cteristic for the
first, action for the second./) If we derive the st, !zn jewels from
The iaoels are (were) stolen we therefore not only derive a less
ambiguous construction from a more ambiguous one but, since the
iewels stolen is supposed to be an intermediate step on the way to
the stolen iewels, we get the illogical sequence 4- characteristic -+
- characteristic ~ + characteristic.
-

W i t h adjectives the contrast - when postposition is possible - is

along somewhat different lines, since the passive voice iz not
involved; instead of characteristic vs. action it is characteristic vs.
occasion. E x a m p l e s with the -able adjectives me most typical: T],e
only river that is navigable is to the north does not tell us whether

x) The relation of the latter to the passive voice is ~vident in the accepta-
bility ordering of The stolen jewels were his, The stMen jewels are his, The
jewels stolen were his, The jewels stolen are his. The la~,t sentence is lowest ou
the scale: jewels stolen p r e t t y clearly refers to one act of stealing, in the past,
which sorts better witl. a following were. Also, for those who distinguish
between burned and bur~,t, the normal positions are ~nown in the examples
The paper b ~ r ~ d wt~s my letter a n d The burnt paper was thrown away.
4 DWIGHT BOLINGER

the temporary states of rivers are referred to ('The only river that
happens to be navigable at the moment'), or the classes of rivers.
But the only river navigable is unambiguously occasion, the only
navigable river unambiguously characteristic. Similarly with Who
were the guilty people?, which characterizes and classifies, vs. Who
were the people guilty?, which relates the guilt to an occasion; and
with The visible stairs were A ldebaran and Sirius, referring to stars
inherently visible (i.e. of a high magnitude), vs. The stars visible
were Aldebaran and Sirius, referring to what could be seen on a
cloudy night. The contrast shows also in the acceptability ordering
of The candidates most active in the campaign were the ones elected
(as a result of their te'nporary activity) and The most active candi-
dates in the campaign were the ones elected," a better context for the
latter is The most active candidates in the campaign were Joe Smith
and William Butler. It sometimes happens that an adjective de-.
veiops two distinct senses related to its positions. Thus The ,man is
responsible is ambiguous as between 'trustworthy' and 'to blame',
but the man responsible is unambiguously 'to blame' and the re-
sponsible man is almost unambiguously 'trustworthy'. Whiskey
straight is a drink, readied for the occasion; straight whiskey is a
product, so characterized by its label.2)
A third reason for skepticism about be predications as a necessary
source for attributives is the more obvious relationship, in many
instances, to predications of other kinds. I do not argue for these
other predications as SYNTACTIC ~ources of attributive adjectives
- only as being more plausible tia,~n be predications. I can think
of the following four:
(1) Adverbial predications from which the advet.~ is recovered
as an adjective. Thus while a daily occu,'rence m a y relate to The
occurrence is daily (even though the latter .~s less frequent), a daily
newspaper seems to relate to The newspaper ".';, ~,,,'~ " d~i!v. A stray
bullet relates not to *The bullet was stray but to The bullet went
astray. She is a constant companion relates to She is constantly a
companion; He is an et~,nal /fiend relates to He is eternally a ]fiend.
He is a / a i r shot could be related to He shoots/airly except that that
sense o f / a i r is not adverbialized; it appears more clearly as His

2) For additional examples see Bolinger, 'Linear modification', P M L A

67 (1952) 1117-1144, esp. pp. 1133-1137.
 ADJECTIVES IN ENGLISH 5

shooting is/air. A big eater relates to something like He eats a lot

(though *He eats big is not used, there is an analogy iv He talks
big). A northern state has only a dubious counterpart in ?The state
is northern, but a good one in The state lies tc the north. A n occasional
sailor strolled by relates to A sailor strolled by occasionally, s) A n
erstwhile/fiend relates to Formerly he was a/fiend. Some attributives
can be accounted for only on the theory that adjectives exercise a
kind of hegemony in mgdification, and tend to crowd out adverbs.
Examples like the following are fairly common: The car lost: complete
control o/itsel] (completely lost control of itself); I an, out o/ all
sympathy with that movement (I'm ali out of sympathy) ; These pans
are slight seconds, an advertisement (are ,dightly seconds); There
is a definite shortage o / g u m (There is definitely" a shortage of gum).
(2) A few relics of the ancient perfect tenses with be: The Indians
are (have) vanished ~ the vanished I n lia~,s; The gue,,:s are (have)
departed ~ the dep~rted guests. 4)
(3) The passive voice. It can be argued that a phrase like siolen
iewels ought to be related most direcCf to be predications like He
re/used to buy the iewels because they were stolen rather than to the
passive. The trouble with this is that countless perfectly good
attributior_s do not have a counterpart be + state construction. We
readily say increased prices but never or almost never These prices
are increased in the sense 'These prices are higher'. With a ]ancied
all/terence we tend to nominalize the predicate if we want it to refer
to ~;tate: The di//erence is a ]ancied one. The depos#ed money and
the re/used invitation have no counterpart stative *The money is
deposited and *The invitation is re/used. A better cas, can be made
for deriving the stative be predicatives from attribu:ives than for

a) This t r a n s f e r of t h e a d v e r b is possible because of w h a t m i g h t be called

s t r o b o s c o p i c ~ir, gulars. I n Occaaionally a sailor strolled by we h a v e a tr,ae
sin~';ular in a sailor: t h e occasionally applies t o each sailor + e v e n t sepa-
rat,;ly: ' I t occasionally h a p p e n e d t h a t a sailor strolled b y ' . B u t in A sailor
strolled by occasionally (now an~ then), a sailor i n t e g r a t e s t h e s-~ilors - it is a
sailor w h o occurs s p o r a d i c a l l y on t h e scene. T h e stroboscopic singular can
be modified b y an a d j e c t i v e referring t o e v e n t s : an occasional sailor, an
in/requent visitor, a sporadic s~ot.
4) T h a t these are relics is a p p a r e n t in t h e impossibility of using close
s y n o n y m s in t h e s a m e way. Disappear is a n e w e r word t h a n vanish: it e n t e r e d
t h e language t o o late for t h e a t t r i b u t i v e c o n v e r s i o n of be p e r f e c t s to t a k e
effect: *the disappeared Indians.
6 DWIGHT BOLINGER

doing the reverse. If we are to derive these from sentences it seems

best to go to the passive voice directly.
(4) Predications from which the verb is recovered as well as its
complements. These are typically c o m p o u n d -ings a n d compound
deponent -eds:
The m a n w d k s slow ~ a slow-walking m a n
The ~ r ! !o,~es home -o a home-loving girl
The scene fills the eye ~ an eye-filling scene
The child behaves b a d l y -+ an ill-behaved child
The womav travels widely ~ a widely-travelled woman
i t would be po,;sible of course to p u t a be predication somewhere
between the ex~ remes here, thus:
The girl loces home -+ The girl is home-loving -+ the home-
loving girl
b u t since only well-established forms like home-loving, eye-/illing,
a n d z~u-oenavea
: " ~-'- .... "~iavor
. . . . . it,
"~ i.e.,
The slouctz chews tobacco ~ *The slouch is tobacco-chewing -+
a tobacco-chewing slouch
The m a n h a t e s women --, *The m a n is w o m a n - h a t i n g --, a
woman-hating man
Your friends write letters -+ *Your friend.,; are letter-writing -+
your letter-writing friends
are not good derivations, as with the passive a b e t t e r case can be
made for deriving the be predications from the a t t r i b u t i v e s - the
compound ,,,~j~u-,~~a'~-*:-'~finds_, its w a y to the predicate o n l y a.fter it has
become entrenched. 5)
Observe now how differently we react to the k i n d of transfor-
m a t i o n represented b y the last set of examples, as compared with
active-passive. I t causes us no surprise when some new verb
appears in the active a n d i m m e d i a t e l y shifts to the passive: They
napalmed the village gives (God help us) The village was napalmed
immediately. B u t the great m a j o r i t y of predications are not
transposable to a t t r i b u t i v e position: a secretary who erases mistakes
is not a mistake-erasing secretary, nor is a wife who wakes her

5) I t goes without saying t h a t these are to be distinguished from the t y p e

unre-cutting tool, r~g-cleaning device, where for is implied: a tool for w~e-
cut~ ing.
 ADJECTIVES IN E N G L I S H

husband :t husband-waking ~vi/e. These must wait the day when we

have some interest in characterizing secretaries as mistake-erasing
or wi~es as husband-waking. They are lexical coinages tied to a
time and a place, not free-flowing syntactical transformations that
move as smoothly in one direction as in the other. This restriction
confirms the existence of a set meaaing for pre-adjunct adjecti-:res,
which I have called 'characterization'. e)

6) Lexical transformations and syntactical transformations I think belong

at different places in t h e ~rammar. The evidence of non-compounded -ings
is equally convincing. For all their tecundity as a source, there are un-
systematic restrictions. Chomsky makes a good case for verbs having the
syntactic feature [ + [ + A b s t r a c t ] . . . - - . . . [ + Animate]], i.e,, verbs allowing
abstract subjects and animate objects, as a prolific source of adjectival
-ings and, where t h e latter are lacking, of adjectives with other variant
suffixes, e.g. sccry for scaring: The vision scared John, a scary vision. B u t
when he says (Aspects, 227) that these verbs seem invataably to have an
.,j . . . . . ~, use, it is easy to find exceptions. The most notable are synonyms
of to anger, including to anger itself.
The remark angered John, *an angering remark
The remark peev,.sd John, *a peeving remark
And so for to mi[], to pique, and to chagrin, though in]uriate and e~asperate
do give -ings. I t might seem that figurativeness could be a central T factor
in a verb like to tickle, in view of
The remark tickled John, *a tickling remark
The remark floored John, *a flooring remark
b u t then we find
The remark flattened John, a flattening remark
with the same meaning as *flooring, just as figurative, and acceptable. Other
examples:
The remark mortified John, a mortiiying remark, BUT
The remark shamed John, *a shaming remark
The remark deceived John, a deceptive remark, BUT
The remark fooled John, *a fooling remark
The remark frightened John, a frightening remark, BUT
The remark par.ticked John, *a panicking remark
The experience shattered Jc~m, a shattering experience, BUT
The experience broke John, *a breaki~.lg experience
The mishap eonh~sed John, a confusing mishap, BUT
The mishap dazed John, *a dazi~.~g mishap
As might be expected, if there is no truly syntactic transforrnational re-
lationship here the semantic features are o~_~en not the same between the
verb and the -ing;
The remark affected John (perhaps angered him)
an affecting remark ('causing pity')
8 DWIGHT BOLINGER

Other possible sources of attributive adjectiw~s leave us in doubt

whether to include a be-predicative stage or not. Take the adjectival
-ed suffix, which can be related to predications with have:
The man has one eye, the one-eved man
The road has crooks, the crooked road
(Tile typical fossilizations of lexical transformations show up here
in the virtual non-use of ,has crooks nowadays)~ One might or might
not include the extra step with be:
I know the man t
The man has one eye j --~ I know the man who has one eye -+
I know the man wlho is one-eyed -~ I know the one-eyed man
It is obvious that the transformations that can fill the attributive
slot have m a n y sources other than simply be predications. Let us
see now whether even among the be predications themseNes all can
be used. The first step wiil be to show how some attributives and
predicatives have restrictions that bear out the semantic label
'characterization' and suggest the need for recognizing two kinds
of be predications.

2. CHARACTERIZATION
The attributive set with the most striking restrictions is that of
the perfect participles. Tile predicative set similarly endowed is
tihat of adjectives that I shall call 'temporary'.

The remark maddened John ('drove him out of his mind')

a maddening remark ('an exasperating remark')
Also there are commonplace -ings whose verbs ,~e n o t used, or are rarely
used, in corresponding predications:
an arresting experience, *The experience arrested J o h n
a scathing contempt, *Her c o n t e m p t scathed Jobu
a fetching look, *Her look fetched J o h n
And there are other restrictions, e.g. the verb should not be specified Nega-
tive:
The remark d i d n ' t faze John, *a fazing (*unfazing) remark. Synonymic
pressure is a factor in the existence of these adjectives. I f t h e y are coinages
rather than syntactic transformations t h e n we will n o t coin unless we
feel some real or capricious need; and if there is already a word t h a t
satisfies us, we a o n ' t need, and we w o n ' t coin. We h a v e bothersome: we
d o n ' t need lazing; we have a debilitating disease: we d o n ' t need ~a weakening
disease.
 ADJECTIVES IN E N G L I S H 9

(1) Perfect participles. The typical perfect participle that can be

used a t t r i b u t i v e l y is one t h a t leaves a mark on something: a dented
/ender, a wrecked train, a smashed table, a bruised cheek, a /rozen
branch, a smudged eyelid. W h e n one scratches one's head the result
is not *a scratched head b u t when one scores a glass sur:face the
result is a scratched sur[ace. Similarly we have labeled goods but not
*sent goods, dented bells but not *rung bells. But 'leaving a m a r k '
is only the unhewn side of characterization, the most obvious
means of stigmatizing a thittg by what appears on its surface.
It is these examples t h a t appear as easily in the predicate as in the
noun phrase: The goods are labeled, The /ender is dented, Your
eydid is smudged. More critical examples are fl-~ose of verbs t h a t
do not leave marks. Take to deposit and to withdraw. Both can
appear as statives in the predicate: The money is already deposited,
The money is already withdrawn. But while we can refer attribu-
tively to deposited money we would seldom if ever say *withdrawn
money. There is some interest in deposited m o n e y because it
contrasts, in our manner of keeping accounts, with invested money
and pocket money. Withdrawing money does not put it in a situ-
ation t h a t interests us - the culture does not recognize any class of
money t h a t can be so characterized. On the other hand Money
withdrawn does not yield interest is normal, and is as close to a
temporal as to an adjective clause: 'money when it is withdrawn';
the money is not characterized. 7)
(2) T e m p o r a r y adjectives. If an adjective names a quality t h a t
is too fleeting to characterize anything, it is restricted (with t h a t
meaning) to predicative, or to post-adjunct, position. The meaning
of ready in The man is ready or The materials ready will be shipped
is excluded from *the ready man (and this has po!.arized ready to
the extent t h a t a ready wit is not easily interconvertible with ?His

7) Characterization b y what is striking and exceptional probably accounts

for the high proportion of negative participles and others with negative
import: undraped figure more frequent than dr,:ped figure, hated man more
frequent than liked man (w,; have to step the latter up a degree: well-liked
man), lost jewels b u t not */ound jewels. Negation also affects the old be
perfects referred to earlier: though to arrive is as old as t9 depart, we have
departed suests b u t not *arrived guests. An oddity like It was decided by a
tossed coin p r o b a b l y reflects the hegemony of aojective modification alread,,
referred to: ' I t was decided b y tossing a coin'.
I0 DWIGHT BOLINGER

w # is ready). Similarly w i t h handy: Are your tools handy? n o r m a l l y

means 'Do you have t h e m conveniently at h a n d ' ? whereas a handy
tool n o r m a l l y means a useful one - a tool can h a r d l y be characterized
b y the fact t h a t it is m o m e n t a r i l y w i t h i n reach. F,r,ush, for 'having
a lot of m o n e y temporarily', is, as we saw earlier, not used in *a
[lush man. One m a y s a y The house was red in the sunset, b u t it would
take a poet to arrest this t e m p o r a r y image a n d say the red house.
Adjectives referring to t e m p o r a r y states of health, sensation, mind,
or spirits are similarly restricted: H o w is your [fiend? - He's fine
(great, swell, zoonder[ul, lousy, dizzy, hot, blue)~ vs. your fine (dizzy)
[riend it~ a different sense. I've never seen a man so sick - he was
positive!y green does not sanction *a green man. All o / a sudden the
girl w a s / a i n t does not sanction *the [aint girl. Did your [fiend like
your pe~qormance? - Yes, he was quite laudatory is unlikely as ?a
laudatory/fiend. Of an individual who takes a m o m e n t ' s liberties
we can say That man was pretty personal, wasn't he?, from which
we cannot get *a personal man.
Restrictions on temporariness show up in a n o t h e r guise" its non-
use in nominalizations. Of a person who for the m o m e n t speaks too
f a i n t l y we can say I can't hear you; you're not loud enough b u t not
"1 can't hear you; you're not a loud enough man. This contrasts with
I can't hire you; you're not clever enough (you're not a clever enough
~ga n ) .
H o w temporary must a t e m p o r a r y adjective be for a t t r i b u t i v e
position to reject it~ There is obviously no measure for this. A
speaker with a mind to exaggerate can do ,m,iost a n y t h i n g . Thus
The chairman was apologetic s) would not ordinarily be reflected by
an attributive even in the t e m p o r a r y c o n t e x t : Was the chairman
apologetic? - ?Yes, we had an apologetic chairman there/or a minute.
B u t hyperbolically, with a few trimmings, Yes, we had a very [red-
/awed and] apologetic chairman there/or a minute (the most apologetic

s) That. 'temporariness' is the cat~dyst here cart be d e m o n s t r a t e d by

comparing an adjective like apologetic or congratulatory with one like critical.
One is apologetic or congratulatory only long enough to express t h e apology
or the congratulations. One m a y be critical indefinitely. So c~itieal person
is common, apologetic person is less so (it m i g h t be ,,aid of a Uriah Heep),
a n d congratulawry person is rare. On the other hand, a message conveying
congratulations remains w h a t it is and can be so che.racterized: a congratu-
latory m*:ssage.
 ADJECTIVES .rN E N G L I S H !!

chairman you ever saw). In answer to Were you sorry,~ one might
hear Boy, I was a sorry man all right, about the sorriest man you
ever saw.
But the temporary adjective is in a weak position for attributive
use, and if anything conspires to weaken it further, attribution is
proportionately more difficult. Conflict of homonyms could be
cited for some of the examples already given (green 'sick', green
'color', fine 'healthy', fine 'good'), and for many more: Your/fiend
is high ('drunk') does not give *your high/riend though your tipsy
lriend is normal. Temporary adjectives like jumpy, downcast, upset,
and ill are possible, though a bit unusual, as predicatives - there is
no conflict. The antonyms present and absent pose an interesting
contrast: your absent/fiend but not *your present/riend. The latter
is exposed to a conflict of homonyms and the former is favored by
being negative (like departed guests vs. *arrived g~ests).
The opposite m a y also occur - a temporary modifier becomes
normal if the situation is such that nouns are distinguished by it.
Adjectives (or adverbs doubling as adjectives or vice versa) referring
to location in space and time in relationship to the speaker or to
some other movable point of reference are noteworthy for their
resistance to attributive position. The phrase the then president
is about as far as English has gone in permitting temporal adverbs
to be used attributively. *The now president is impossible, sa) and
here and there are inadmissible. The adjective nearby can be used
of something stable enough to preempt - location - a nearby
buiMmg, a nearby group - but not -f something that m a y con-
ceivably move off the next moment: *a nearby man, *a nearby bus.
Near,/at, and close are also highly restricted. The predications The
man is clause, The man is near, You're too Jar are possible, but the
corresponding attributions *the close man, *the near man, *the jar
figure are hardly acceptable. Similarly the left dog and the right dog;
we prefer the dog on the left. (My ]viend is close versus my close/fiend
reveal the familiar change of meaning.) The near side and the jar
side or the near corner and the/at corner, or the hither side and the yon
side, are significant contrasting dimensions of these nouns, like
inside and outside or top side and boa,tom side. A similar contrast

as) B u t almost as I write this an enterprising beverage c o m p a n y adver-

tises T h e n o w taste o[ Tab.
12 DWIGHT B O L I N G E R

enables the comparative to be used with close, near, and /at," the
closer man, the nearer man, the [arther figure.
A t e m p o r a r y adjective m a y be cemented in place b y a c o n t c x |
t h a t is equivalent to a predication. A loose coat normally means a
loose-fitting coat, one so characterized; b u t loose can also mea~,
'unfastened', :rod in this sense of temporariness to say You'd better
button your lo, ,se coat is a bit inappropriate - we are more likely to
arrange the context as if to assume t h a t a predication has just been
made: You'd better button that loose coat o/yours be~ore you catch
your death, o/cold ( = Your coat is loose + You'd better button it).
More examples will be given later of a t t r i b u t i v e s t h a t are built
on p r e d i c a t i v e s - explicit predicatives, predicatives-in-discourse -
in this way. Loose rope requires no predisposing context.
One set of temporary adjectives is distinguished formally: those
having the prefix a-. These have been restricted to predicative and
dj . ~ . ~ , o z W..J.~...FZA O ~ J
l..13t LIII.~lZ by " " by
O , IJL V V~L O J . O b &

sense of temporariness (the two factors are related, of course): a

house afire, a man asleep, arms akimbo, The people were alive, agog,
etc. Some are gradually edging their w a y into attributive position:
a sensitive and aware audience; 9) the only regular-season games they
ever sold to tdevision were the away games they played against each
other. IQ) Away games contrasts with home games; a recognized class
of games is given its a t t r i b u t i v e characterization.
Adjectives with complements of their own m a y of course be
excluded by that fact from attributive position:
The man was loth to speak, *the loth m a n
The man is averse to speaking, *the averse m a n
But this has nothing to do with temporariness; the s y n o n y m s
unwilling and reluctant can readily be attributive.

3. T w o KINDS OF PREDICATION
We have seen numerous examples of a m b i g u i t y with be pre-
dications which are cleared up by pre-adjunct or post-adjunct
position of the adjective: the ambiguous The river is navigable vs.
the unambiguous river navigable and navigable river, for instance.
We have also seen interconvertible predicatives and attributives

9) Harper's Magazine, Nov., 1963, p. 54[2.

lo) Saturday Evening _Post, May 2, 1964, p- 73.
 ADJECTIVES iN E N G L I S H 13

hke The girl is [odish - the ]oolish girl and predicative~ t h a t are not
convertible iike The girl is/aint, *the/aint girl, where the same is
appears in both. Gr is it the same? Is it possible to project the
river navigable vs. navigable river contrast onto the verb and come
out with two different kinds of be predications ?
A look at other verbs m a y help. Except for those with be, pre-
dications t h a t can appear as a t t r i b a t i v e s show occasional formal
differences from the ones t h a t c a n n o t : the dinosaurs ate the /i~h
does not give *fish-eating dinosaurs, "n) The dinosaurs ate fish does,
in two of its three senses ('The animals known as dinosaurs were
eaters of fish', 'The p a ~ i c u l a r dinosaurs were eaters of fish' - non-
restrictive and restrictive modification respectively; But not 'The
particular dinosaurs dined off fish on a particular occasion or
occasions'). In so far as predications show a formal difference
between customary action and non-customary action, this differ-
ence is a clue to the ones t h a t can be used att~'ibutively. The man
broke a leg does not give *a leg-breaking man," a bastard who ran
rum (not the rum) during Prohibition was a rum-running bastard. 12)
W h a t one does customarily is useful for characterization.
W i t h be + predicate adjective we do not find formal dilferences
between customary a n d non-customary: The man was ,mad readily
gives the mad man (and thence the madman) in the sense 'insane',
but there is no formal difference between this and The man was ma.4
in the sense 'The m a n was temporarily m a d with anger' or just
'The m a n was m a d ' in the current sen~ ' transferred from the
latter. (We see from this w h y mad in the. sense 'angry' has never
m a d e it to a t t r i b u t i v e position). B u t on the strength of the formal
differences elsewhere, one can argue for two kinds of be predications:
the t y p e The girl was ]aint, which analogizes with The man broke a
leg and does not yield attributives, and the type The girl was
[oolish, which analogizes with The girl owned property and does
yield attributives. The latter, if the analogy were carried cut
formally in attributives, would give /oolish-being girl to parallel

xl) This is a good as place as any to enter a caveat against derivi~ag /'~sh-
eating from is eating lish. Fish-eating from the aspectual standpoint is tne
opposite of eating 1ish - the one is charac,~eristic, the other is temporary
A form such as house-hunting (He is house-hunting) is highly exceptional.
13) Since this is lexical transformation, it is not free; rum-running bastard
is normal, but rum-running man would not be used - rum-runner is available.
14 D~;'IGHT B O I I N G E R

property-owning .~iri. To dist;.aguish the two aspects - t e m p o r a r y

and non-temporary or customary - I shall ,,11 the two be predica-
tions bet~mp and bentemp, i t is bentemp predications t h a t underlie
most attributives - ~ t h simple adjectives (others, as we have seen,
relate to adverbial, passive, etc. predications). Unlike other verbs,
the being of [oolish-being girl, red-being house, tall-being tree is
obligatorily deleted.
Copulative verbs t h a t do retain the -ing with their attributives
cor.firm the analogy between be and other verbs b y showing the
same restrictions with t e m p o r a r y adjcctiw~s a.~, be. T h a s The tool
looked handy can mean either ' r e a u y to h a n d ' or t~Ls~fm, but
handy-looking tool can mean only 'useful'. *Faint-loo'i~,,g girl is as
unlikely as *]aint girl, b u t ]oolish-looking girl comes a~ readily as
[oolish girl.
The appeal to c u s t n m a r y aspect accounts for some of the re-
strictions on a t t r i b u t i v e ~dio~t;,,.,~ 1~ n,,~ ¢Le, e o,0 ,~¢1,~,-o

4. REFERE NCE-MODI~TICATION
A derivation like the following seems acceptable:
I saw a m a n /
The man was h u n g r y i --* I saw a h u n g r y m a n
B u t this o,~e is less so"
I saw a s t u d e n t i
The student was eager / -+ I saw a n eager student

13) The semantic label 'customary' overstates the case, which is w h y I

prefer ben temp LO becus. M a n y instances of be predications and of other
predicatioas t h a t can be viewed as underlying at'tributives do n o t accord
fully with 'customary' in t h e sense in which it is applied to hhe aspects of
verbs. I t is far-fetched to say tizat I am visiting a s i c k / f i e n d has to do with
someone customarily sick. The point is t h a t he is not m o m e n t a r i l y sick
- betemp is the m a r k e d t e r m of the opposition. The paxallel with customary
aspect in predications with other verbs does n o t seem to be amLaccident, and
there the case for ' c u s t o m a r y ' is stronger. We do find exceptions, e.g. See
that beer-guzzling charaae~, over at the bar?, where the person referred to does
not nece,~sarily guzzle be,~r habitually, b u t this is best viewed as a figure of
speech. T'ake away the sLrong feelings t h a t lead to name-calling metaphors
and one is less likely to find -ings used in this w a y - it would be unusual to
hear See that book-readilsg man over at the library table? The -er agentives,
whose ties with c u s t o m a r y aspect are unmistakable (a b u y e r is one who
makes a business of buying, not one who b u y s just once), are also used
figuratively in the same w a y : See that beev-gu:zler over at the bar?
 ADJECTIVES IN G N G L I S H !5

Its f~z.ling comes clearer if both predications are with be:

The boy is a student }
The student is eager J --, The boy i~ aii eager s t , d e r t

It has a sort of undistributed middle: The student is eager uses

student in a different sease, as a designation of an individual, and
amount~ to saying T'.,nc boy "~" ,,,g,'r
. . . . . . . . An
. . ,,ag~r studc~t, however,
suggests someone who is eager qua student. The hrst ~s referent-
modification, the second is refer'ence.-modification.
Predications, since they modify the referent rather than the
reference of the noun, thus turn out to be unsatisfactory sources for
m a n y attributive adjectives. An extreme case is one like the follow-
ing:
Hcitry is a policeman [
*The policeman is rural j .+ * H e n r y is a policeman who

is rural --, H e n r y is a rural policeman

The u n g r a m m a t i c a l i t y of *The policeman is rural illustrates the
divergent restrictions t h a t apply to the two uses of the noun, as
subject of a v, cdication and as part of a noun phrase. In the first,
which is referent-modification, it is restricted b y the CATEGORY
of the noun; in the second it m a y or m a y 1~ot be. R~rai is not used
with h u m a n subjects, though it m a y appl'.' to an abstraction such
as 'status of being a policeman'. We see this ..-'ore clearly with an
ambiguous adjective: a regul,r policem,~n vs. The police~nan is
regular. The attributive refers to regularity qua policeman; the
predicative to regularity qua h u m a n being - it m a y mean t h a t he
brushes his teeth regularly or has regular bowel movements. The
difference between reference..tnodification and referent-modifi-
cation often selects different meanings of the adjective, e.g. i n a
criminal lawyer vs. The lawyer is criminal. Lawyers are classed as
criminal, civil, etc. ; an individual m a y be classed as criminal or as
law-abiding.
If we are determined to do so, we can manage some k i r d oi
predicative foundation for reference modification. One device is
with a predication t h a t amounts to a definition: Lawyers are
criminal, civil, constitutional, etc., or A lawyer can be criminal, civil,
constitutional, etc.. This limits the semantic range of lawyer to the
reference of the word, but in a transformation it begs the question:
16 DWIGHT BOLINGER

J o h n is a lawyer /
Lawyers are criminal, civil, etc. ~ ~ ?
This has nowhere to go; one must fall back on a nominalization:
- J o h n is a lawyer /
The lawyer is a criminal lawyer j -~ J o h n is a criminal lawyer

The m a t t e r of nominalization is significant also in those few cases

where a reference modifier CAS appear in the predicate in i,ormal
language. Thus along with a commercial agent one might find This
agent is commerci:d, or even Is .Ioh,, an insurance agent? - No, he's
~,,,,,,........ ~. ,,~,~ since in trade lingo the l a t t e r can be turned around
with the result that there is no adjective at all - Is John a commercial
agent? - No, he's insurance - it appears t h a t commercial is a stand-in
for a nominal. 13s) Nominalizations readily o c c m - i n sentences of
this type without producing the sensation of excessive r e d u n d a n c y
that t h e y do elsewhere"
The agents in this building are mostly theatrical (ones, agents)
The novelists we studied were mostly regional (ones, novelists)
The addresses in these volumes are r~ostly presidential (ones,
adresses)
The engineers in this department are all mechanical (engineers)
This calculus is integral (integral calculus, the integral kind)
The New York theater is mostly legitimate (theater)
These nouns are common (n,~uns)14
Co atrast these with
The addresses in this volume arc iaostly interesting (?ones)
The giri~ in this dorm are all p r e t t y (?ones)
where nomir, alizing is out of order unless the ground is carefully
laid (e.g. by a mention of contrasting cases - homely girls, for
in~:tance ~ in the immediate context).
50 it seems that the attributive use - which of course is the one
embodied in a nominaiization is the primitive one and the pre-
-

:-~'-) Sayo Yotsukura in a projected paper discusses the t y p e I ' m the fifty
cements, identifying oneself as the pevsor, to w h o m fifty cents is due. This I
fix d normal in English.
t4} The unacceptability of such t..edications in a n y but a narrowly
de :ining-and-classffying c o n t e x t ca~ be r,oted in * I a m going to study calculus
th, t is integral: * We want to con:sult agents who are theatrical.
 ADJECTIYES IN E N G L I S H 17

dicative is derived from it, rather than the reverse. To say Lawyers
are criminal, civil, etc., or Lawyers are criminal as well as civil,
etc. is a trick for saying The word lawyer can be modi/ied by criminal,
civil, etc.
The logical conclusion is to provide for this additional determi-
nation of the noun phrase - reference-modification - in the kernel:
Itenry is a kind of policeman
kind of ~ rural, urban, regular, special . . .
The party was a kind of undertaking
kind of ~ joint, individual, collective...
Non-predicative generation of this kind of determination is sup-
ported by a comparison that was made earlier between be and look.
This was according to the following scheme:
The gir! is foolish } { foolish-being girl -~
The girl looks foolish ~ foolish-looking girl
foolish girl
where the predicative origin of /oolish-looking is apparent but is
obscured vith be because of the rule deleting being. Applying the
same test here we find that the adjectives that require an attributive
slot and are not to be traced to any predication are exactly the
ones that do not admit of a compound attributive modifier with
-looking:
*rural-looking policeman
*mere-looking kid
*old-logking school (in sense 'former s,'~hool )
*distant-looking cousin
*personal-leoking friend
*tactile-looking organ
Attributive adjectives that can be traced to predications do admit
of-looking (apart from situations where some other copulative
may be called for, e.g. melodious-sounding singer, crooked-acting
cousin):
drowsy-looking policeman
friendly-looking soldier
useful-looking tool
hard-looking surface
deadly-looking cobra
18 DWIGHT BOLINGER

The typical reference-only modifier is an adjective that does not

take comparison: These nouns are common (ones), *These nouns are
very common (nouns). But many reference-modifiers do admit
comparison. What is generated in the kernel is not a list of adjec-
tives but a reference-modifying slot, occupied by adjectives some
of which are free to appear in the predicate, others not, but all of
'which, when in attributive position become modifiers of the refer-
ence system of the noun, not of any particular referent directly.
Thus a crooked lawye~ is one who is crooked in his practise (and
this m a y take comparison: a very crooked lawyer), whereas The
lawyer is crooked may refer to an individual's personal dealings. The
reference system grips~ the attributive adjective more tightly than
the predicative adjective. Thus The contract is good and The ,marriage
is good can both mear~ 'binding'; but whereas 'binding' i~ germane
to contracts, other meanings of good are more germane to marriages,
and gc,od contract can mean 'binding contract' whereas good marriage
is unlikely to mean 'b:nding marriage'.
An attempt to subc!assify reference-only adjectives is a :~tudy in
itself, but here are some noteworthy types that show especially
well the kinship to delerminers:
(1) Adjectives that identify the reference of the noun with
itself. They tell us t~at the noun means what it says. There is
typically an indefinite determiner alongside:
Hie is a true poet (truly a poet)~; *The poet is true
Hie is a regular champion; *The champion is regular
I-lie is a sheer fraud; *The fraud is sheer
Other examples: plai~,; /ool, arrant rogue, u'der incompetent, per/ect
ass, pure nitwit, unadu]2er,zted ~ackass, mere kid, ordinary policeman,
common soldier, real /fiend, unmitigated liar, total stranger, actual
/,~t, i~onest quart, straight whiskey.15)

15) I t goes without saying t h a t some of ~hese adjectives m a y be used

predicatively in other ftmctions. We have already noted this with This
whiskey is straight, a referent-modification in ".-hic~ the subject is a particular
whiskey drink about which the, question ha., been raised as to whether it
is mixed. One would not bold up a sealed bottle and say This whiskey is
~:traight - rather, This is straight whiskey.
The relationship of such adjectives as me~e to the determiners has been
pointed out b y Sandra Annear, Language 40 1964) 39.
 ADJECTIVES IN E N G L I S H 19

(2) Adjectiw:s that might be called intensifiers of the determiner

the. As we look over this set we recognize that all of them carry
meanings that are often expressed by accenting the word the,
though doing this is more ambiguous. I distinguish three subsets:
(a) The in the sense 'already determined':
He is the very man (precisely the man) I was looking for;
*The man is very
This is the particular spot I me:tnt; *The spot is particular
That is the precise reason I refused; *The reason is precise
For the second of these examples compare This is thd spot, with the
accented. To this subset belong also same,, sel]same, and (in the
relevant sense) identical, exact, and speci]ic (also di][erent, as a case
of not-the).
(b) The in the sense of importance:
.Look for their main faults; *Their faults are main
He was named first citizen; *The citizen was first
He is our prime suspect; *The su:;pect is prime
They were the sole (only) survivors; *The survivors were sole
To this subset belong also principal, chie], and topmost. Note again
the equivalence between He is the prime suspect and He is thd
suspect.
(c) The in the sense 'recognized':
He is the lawful heir; *The heir ,~ lawful
He is the rightful owner; *The ~ ,,'her is rightful
,Mfonso is the trae khxg; * I'he kirg is true
This is the right (wrong) book; *The book is right (wrong)
True has a double entry and same of the adjectives in this subset
have other functions, e.g. a law[ul act,. This man is right ('acceptabk;')
/or the job, etc. Once more compare He is the right/ul owner wffh
He is thd owner, Though adjectival predicates are ruled out, nomi-
na]ized ones are normal in all three subsets:
That man is the very one
Those faults are the main ones
That heir is the lawful one
(3) Adjectives t h a t seem to qualify the tense of the verb in
some underlying structure. Just as the preceding set can be regarded
20 DWIGHT BOLINGER

as adverbial modifiers of the (the very man = precisely the man),

so this set can be regarded as a tense modifier t h a t has been shifted
onto the noun:
He's the future king; *The king is future
This hor'~se is a sure winner; *This winner is sure
This is m y old school; *The school is old
Kennedy is our late President; *The president is late
Other members of the set: erstwhile, quondam, whilom, [ormer,
coming, budding, and p r o b a b l y also putative, possible, pro~,able, and
likely. Since the adjective represents an adverbial on the verb, it
is not surprising t h a t when the noun is an event noun, predications
are grammatical:
Fut~.re events cast their shado~ ; These events are future
I can promist ~you a sure win; This win is sure

di,~cussed wel e old re)

5. REFERENT MODIFIC~ TION

The apparent accepta'kility of

I saw a man /
The m a n was h u n g r y j + I saw a h u n g r y m a n

and the unacceptability of

I saw a lawyer [
*The lawyer was criminal j -~ I saw a criminal lawyer

I judge to b e due to the semantically reduced status of man: it

represen~ts little more t h a n the category H u m a n , so t h a t reference
modification, which applies to the semantic range of the word as
a word, and referent modification, which applies to the category
(in this case Human), coincide. For the same reason one would
not get, as a modification of the reply in Who'll be our next chair-
.man? - Your brother would be excellent the sentence *He is an excellent
brother (tor the purpose), b u t could readily get He is an excellea;

16) Compare also t a e position of only a n d just relative to the rest of the
n o u n phrase, suggest ~ag an adverb: only (just) a boy -~ a mere boy.
 ADJECTIVES IN E N G L I S H 21

man (/or the purpose). ]'hen the doubtfulness of

I saw a student /
The student was eager I ~ I saw an eager student

is due to the ambig~fity of The student was eager. As a H u m a n

noun, student allows (4! eager in a n y reh;vant sense, including 'eager
qua student' but als:. including 'eager to be off', 'eager to hear
the latest racing new:;', or vchatever. Eager student, on the other
hand, normally means 'eager qua s t u d e n t ' - reference modification.
Yet there are times when a noun t h a t is not semantically bare,
like man, takes modification t h a t is clearly not reference modi-
fication, e.g. A drowsy policeman sat at the door, supposedly to
guard against intruders. It is clear t h a t drowsy policeman does not
mean 'drowsy qua policeman'. E v i d e n t l y there is a way for referent-
modification to become attributive even with semantically rich
nouns ,.~,~ lm~,,emun. And given the referent-modification of
1~1.- &^l" . . . . . . .

predicatives, it would be good to have predication as a source, to

derive drowsy policeman from policeman is drowsy, or something
like it.
Since referent-modification goes to the category of the noun, a
predication that is to serve as the source ought not to specify
semantic features for the subject. I suggest therefore an existence
predication like the following:
There is a person such t h a i he is a policeman ]
There is a person such tha~ he i:~ drowsy [ --~
There is a person such tt at he is a policeman and drowsy
There is a person such t h a t he is a drowsy policeman
(where referentl = referentg, ar:d the is is bentemp). Or, using some
more c o m e n i e n t designation for the subject,
H e n r y is a police man I
H e n r y is drowsy l -+ 2Ienry is a policeman
and drowsy -~ Henry is a drowsy policeman
In other words, a conjunction rather than a relative clause. The
conjunction expresses the independent relationship of adjective
and noun: his being drowsy has nothing to do with his being a
policeman. It is not t h a t pdicema:e who is drowsy necessarily implies
a n y dependence - it is simply neutral when what we need is to
22 DWIGHT BOLINGER

specify independence27) To try to express this iv. the usual way,

"' Henry is a policeman /
The policednan is drowsy j -+ H e n r y is a policeman
who is drowsy -~ Henry is a drowsy policeman
conceals the fact that The policeman is drowsy really means The
policeman Henry is drowsy -policeman is a stand-in for the H u m a n
referent.
Reference-modification as one of the determiners, and referent-
modification as the product of a conjunztion, are evident in the
following contrasts:
Henry is a distant cousin
?Henry is a crooked cousin
Henry is a crooked cousin of mine
Henry is a crooked lawyer
Distance in cousins and crookedness in lawyers are part of the
tC,t~J.~;n~t~ a VaL~illo Lu~.ro~, nouns. ~.v a~,~.A,.,.,a v~

reference system of cousin. To use crooked with cousin as part of

an is a predicate calls for a predetermined referent, which is sup-
pried by a . . . o] mi~e: 'There is a cousin of mine and incidentally
he is crooked'. A similar example with au inanimate noun:
What's that piece of paper? - ?It's a gracious letter
What's that piece of paper? - It's a gracious letter that I've
just received
What's that piece of paper ? - It's a personal letter
Personalness is part of the reference system of letter; graciousness
is not. But if a particular referent is determined, as by that I have
just received, referent-modification can appear attributively: 'It's
a letter that I've just received, and incidenta!ly it is gracious'.
Similarly with
Who is John Smiith? - He's a famous lawyer (popular enter-
tainer, respected minister, much sought-after speaker,
greatly admired doctor, benevolent dictator)
Who is John Smith ? - *He's a nice cop
Who is John Smith ? - He's a nice cop that I know

XT) In a sense it does specify it here, since the who of who is drowsy is
pure category: who = Relhumsn. B u t this does n o t help with I n a n i m a t e
antecedents, where which does not distinguish category from non-category.
 ADJECTIVES IN ENGLISH 23

Nice cop requires, in an ,s a predicate, something to indicate a

particular referent.
I n a n i m a t e nouns show the contrasts less sharply t h a n H u m a n
nouns, but t h e y are presen: nevertheless. For example This book is
short is more apt to be resl:ricted only as to the category Material
Object, hence to allow of a n y kind of shortness, including length
in inches as well as length in pages, whereas a short book is more
apt to mean shortness qua book, hence n u m b e r of pages. To gener-
alize the existence predication to include N o n - h u m a n nouns merely
calls for There is an X such that rather t h a n There is a person such
that.
H a v i n g provided for the two extremes of referent-modification
and reference-modification, we can now assess the blending that
f r e q u e n t l y - p e r h a p s more often t h a n n o t - takes place between the
two. While a drowsy policeman scarcely refers to someone who is
drowsy qua policeman, a ]riendly policeman does suggest friendliness
in the way in which policemen manifest it - courtesy in dealing
with the public, helpfulness to old ladies, good fellowship along the
beat. 1'~) A happy agent suggests one who is h a p p y about his work.
The agent is happy allows of any m e a n i n g of happy t h a t can go
with a H u m a n subject - wearing a h a p p y expression on his face
as readily as h a p p y qua agent. B l e n d i n g is to be expected, giwm
the nature of nouns, which both name classes and designate indi-
viduals, often doing bo~h things at once. In their capacity as
namers of classes t h e y take reference-modification; as designators
of individuals t h e y take referent-m~dification.

la) There is a certain amount of semantic adjustment m the corresponding

.predication too: The policeman is [viendly. All that this means is that ad-
justment even across a conjunction - policeman and friendly - i s to be
expected. The friendliness referred to is more likely than not to be friendli-
ness in the way policemen arz usually friendly. But the predication is per-
fectly free to take on other meanings of/riendly, whereas/riendly policeman
is much less so: That policeman is certainly/riendly couht refer to a neighbor
who manifests neighborliness and only happens; to be a policeman - policeman
does no more than identify the person about whom we, are speaking. This is
why I think that the semantic adjustment should lye sought within the
attribution and not within a prior predication.
24 DWIGHT BOLINGER

6. ESTABLISHMENT IN DISCOURSE
Transformational relationships presumably are paradigmatic.
They define structures in terms of structures, the relationship
between them being bi-directional and the notion of 'transfor-
mational history', i.e. of starting point and destination, ihaving
nothing to do with events in discourse. There is thus a paradigmatic
and not a syntagmatic relationship between active and passive,
for example. The active is related to the passive, but that does
not mean that anyone has to say or to pretend to have said or
heard an active before he can say a passive. A speaker Call enter
a room and say, out of the blue, D i d you see that boy? He was just
run over by a t r u c k / w i t h o u t a preceding A truck ran over the boy
having been said or assumed.
Despite the irrelevance of syntagmatic history to transformations,
it is relev,lnt to certain things that happen to the modification of
nouns by adjectives. Establishment in discourse is a matter of
usage which helps to determine how the grammatical resources of
referent-modification and reference-modification can be tapped.
Take these examples:
(l) Short book, as we have seen, is normally taken as 'short qua
book', short in number of pages, lines, paragraphs, etc. One would
be unlikely to say, on moving into an apartment and distributing
articles among the shelves along the walls, This short book is about
right/or that low shelf, though This book is short - it's about right
/or that low shelf would be normal enough. Later, H a n d me that
short book you had would follow from the prior predication. On the
other hand, This short stick is about right to prop u~p the shelf would
cause no surprise. Stick selects from the semantic range of short
only on the basis of Material Object. But with sharp the situation
is different: Get me some sharp sticks to stake up these plants would
refer to sharpened sticks, a kind of stick --sharpness is part of the
reference system of stick. To refer to sticks th3.t happen to be
splintered and dangerous to handle - i.e., sharp in another sense
than 'intentionally sharpened' - one would no1 say *Look out,
those are sharp sticks - rather, Look out, those stick~ are sharp. Then
having established this in discourse one m a y lax:er say I got my
hands full o/ splinters from those sharp sticks. W~ith bits of glass
nothing needs to be established in discourse: Look out - those are
sharp bits o[' glass.
 ADJECTIVES IN E N G L I S H 25

(2) Consider short when the noun is Human. A short policeman

is normal reference-modification" physical size is relevant to being
a policeman. But the predication The policeman was short can
include any meaning of short that is congruent with a H u m a n
subject, e.g. he was curt or temporarily out of money. Having
established this in discourse, A short policeman is resented by people,
meaning curt, is possible. (Short policeman meaning 'temporarily
out of money' would probably be excluded as a temporary adjec-
tive).
(3) Older applied to such a word as daughter is part of the refer-
ence system: older daughter has 'a set meaning in familial relation-
ships. The daughter is older, on the other hand, can mean 'than I
expected her to be, than the cousin', etc. Establishing this in
discourse makes older daughter possible in a context like The two
daughters o~ our two /riends' /amilies ~were there, one older than the
other," the older daughter ( = the •lcler girl) was ot,;.....
(4) You'd better button your loose coat we saw was a bit odd - since
loose coat suggests loose qua coat, a coat of the loose-fitting type -
unless cast in a form that implies a prior predication: You'd better
button that loose ,'oat o/yours ( = Your coat is loose + You'd better
button it) or You'd better button your loose coat there ( = Look there,
your coat is loose + You'd better button it).
Prior predication, establishment in discourse, thus enables a
speaker to transfer a referent modification to attributive position
when otherwise an ambiguity or anomaly would result because of
the more usual reference modific.~tion of the adjective. By this
process a criminal lawyer may refer to someone who is himself
criminal, not necessarily: to someone who practises criminal law:
X is a lawyer l
X is criminal / ~ X is a lawyer and criminal -+
X is a criminal lawyer
- predicative generation rather than attributive.
If the adjective is one whose semantic features can apply only to
the category, no prior predication is necessary. Thus the example
See that drowsy policeman over there? needs no groundwork since
drowsy modifies the category H u m a n and is irrelevant to the
reference system of police~nan. See that crimina! lawyer over there?
would definitely need groundwork if criminal -= 'evildoing'.
26 DWIGHT BOLINGER

Since establishment in discourse is just a m a t t e r of usage, it

might seem to have little interest in a discussion of grammar. But
I suspect t h a t it has influenced our g r a m m a t i c a l thinking. We
noted earlier t h a t
I saw a man /
The m a n was h u n g r y j -+ I saw a h u n g r y m a n

which is the familiar transformational derivation of an a t t r i b u t i v e

adjective, seemed possible because man represents little more t h a n
the category Human. There are just enough instances of such
semantically reduced nouns lying around to m a k e t h a t transfor-
m a t i o n look plausible. Now we find a n o t h e r type that reinforces it:
I saw a cobra /
Cobras are deadly j -+ I saw a deadly cobra

- a classical syllogism, which happens to be the most striking

illustration of its broader class, which is simply t h a t of non-re-
mourners
strictive -- - j ' r ___ in general:
Look at the P a c i f i c /
The Pacific is wide ] ~ Look at the wide Pacific

To both Cobras are deadly and The Pacific is wide we could add as
everybody knows. But common knowledge is not essential:
Our secretary is taking dictation /
Our secretary is beautiful ] -+ Our beautiful
secretary is taking dictation
in which Our secretary is beautijful is k n o w n to the interlocutors,
and functions in the same way as the examples of common know-
ledge. In other words, if we assume t h a t Our secretary is beauti/ul
has been established in discourse, all three of these a p p a r e n t l y valid
derivations involve non-restrictive clauses: I sa~!: a cobra, which is
deadly; Look at the Pacific, which is wide; Ou~ secretary, who is
beauti/ul, is taking dictation.
All of these are to be based, I think, not on the classical transfor-
m a t i o n but on the same kind of predicative-to-attributive t h a t was
posited for the drowsy policeman type. The difference here is t h a t
there is a prior predication which imposes a relationship of 'if-then'
on the source sentences. W h e t h e r the prior predication is something
a,.tually said, or known to be universally true, or just assumed,
 ADJECTIVES IN E N G L I S H 27

makes no difference; the adjective has a predicative source and

the transformation is a conjunction:
There is an X such that it is a cobra
[From the prior predication it follow,; that] -+ There
There is an X such that it is deadly
is an X such t h a t it is a cobra and [therefore]
deadly ~ There is an X such t h a t it is a deadly cobra
(Sentences of the 'everybody knows' type occm more n a t u r a l l y
- less poetically or less epithetically - in contexts where the nouns
are in contrast, e.g. The tree world mani/es ts ~'reat variety, all ~he
way /tom evergreen hemlocks to deciduous cottor,;woods).
Non-restrictive modification with its prior predicatiom; has been,
I think, one of the lures toward a clause transformation (cobras,
which are deadly) rather than the conjunction transformation
advocated here. But it also poses a problem for the difference
between reference-modification and referent-modification" why not.
regard deadly cobra as 'deadly qua cobra' or lousy ~o1~ as 'lousy
qua sop'? i t would seem plausible, for people who t h i n k t h a t all
cops are lousy, to incorporate lousy in the reference .,~ystem of
cop. I see two reasons for avoiding this. One is t h a t a](1 sorts of
disputed value j u d g m e n t s would have to be added to reference
systems (He's ~ust a petti/ogging i'awyer would incorporate petti-
]ogging in the reference system of lawyer). The othe~r is t h a t tile
identical transformation applies to the beaut2i/ul secretary type in
which the prior predication is indi, ddual, not universal, i.e.,
There is an X such t h a t she is our secretary
[From the prior predication it follows that] ~ There J

There is an X such t h a t she is beautiful

is an X such t h a t she is our secretary and [therefore] beauti-
ful --, There is an X such t h a t she is our beautiful secretary
- in which beauti/ul is obviously not a part of the reference system
of secretary. It is best to think of deadly cobra as a prior cobra is
deadly, paralleling secretary is beauti/ul, but infinitely ,extended.

7. AGENTIVE NOUNS AND LEXICAL DERIVATION

In the examp!cs thus iar I have largely avoided agentive nouns.

These appear superficially to be so definite a class and to reflect
28 DWIGHT BOLINGER

so strongly on the predicative-attributive opposition that 1:hey

~ieserve separate treatment. We shail consider whether they do.
There are two typical situations. In the first, an anomaly re~,.ults
when the reference-modifier i's made a referent-modifier, mainly
as a result of the H u m a n category of the subject:
a subterranean explorer; *The explorer is subterranean
a microscopic donor; *The donor is raicroscopic
an electrical worker; *The worker is electrical
a regional novelist; ?The novelist is regional
Where the adjective can modify H u m a n referents, it m a y go in the
predicate:
a stingy donor; The donor is stingy
The give and take between the two sources of attributives raakes
it possible to get a predication out of an attribution. Thus I wou'[d
not be surprised at
an astronomical bar; That 1tar is astronomical
with a figurative extension of the adjective. If the subjec~L is
Inanimate, then some restrictions on predication fall:
an electrical connector; The connector is electrical.
In the second situation, with good-bad adjectives, either position
is normal:
a masterful player; The player is masterful
a first-rate cook; The ::ook is first-rate
a skilled draftsman; The draftsman is skilled
But there may be a conflic~ of homonyms (or - ~¢hat amounts to
the same thing - a varying selection from two widely separa~.ed
parts of the semantic range of the adjective):
a poor liar; The liar is poor
a poor loser; The loser is poor
a beautiful singer; The singer is beautiful
The adjectives poor and beauti/ul have been heavily polarized with
H u m a n referents. 19) It is possible, but unusual, to have a sentence

19) ~¥ith poor the polarization is not so e x t r e m e as to exclude all b u t the

meaning 'impoverished' in th,~ predicate, if t h e reference of the subject
a d m i ~ the notion of 'skill' clearly enough, W e see here the influence of
 ADJECTIVES IN E N G L I S H 29

like That singer is bea.,,~ti/ul referring to the singer's voice. But there
is no problem with
a melodious singer; The singer is melodious.
Mdodious is univocal and not anomalous ,~dth H u m a n referents.
Does the fact that a noun is agentive have any bearing on the
attributive-predicative problem? The particular adjectives that
can be used are definitely affected by the various sources of attri-
butive phrases containing agentive nouns. Some appear to have
got their sta::t from certain possibilities of phrase structure that
would be impossible in predicative position; in the following (as
above with .~ubterranean exploration -~ s,,~bterranean explorer) the
adjective modifies an Inanimate noun:
electrical worker regional novelist public officeholder

Others invoJve adverbs. These obviously cannot be made to depend

on be predicw:ions of the traditional sort:
*The worker is hard -+ He's a hard worker
*The fighters are clean -~ They are clean fighters
But adverbial sources do seem indicated,
He works hard He's a hard worker
They fight clean(ly) They are clean fighters
He bids low He's a low bidder
She cooks well -+ She's a good cook
He visits occasionally He's an occasional visitor
especially if we allow the adverbial pattern to spread over con-
structions that never actually manifest it by grammatical sentences
or if the adverbial sentences are less grammatical than the result:
He holds large amounts of stock --~ *He stock-holds lazge
(in-a-large-way) -+ He's a large stockholder
a t t r i b u t i v e on predicative: reference-modification becomes a factor in
selecting the right meat~ing of an ambiguous predicate adjective. Thus while
He is a poor liar is an incongruous m a t c h for The liar is poor, £'he ,is a poor
typist m a y give, if not readily ?The typist is poor, .at least rather readily That
typist is pretty poor.
30 DWIGHT BOLINGER

He gives low grades -~ He grades low -~ He's a low grader

She ealCs tiny a m o u n t s --~ She eats tinily --~ She's a t i n y eater
He burgles professionally ~ He's a professional burglar
The last example is instructive in t h a t it shows the process in
reverse" burgle is a back formation. It begins to appear t h a t we are
again in the gray area of the type A bullet went astray -~ a stray
bullet, i.e., of transformations t h a t belong in the lexicon rather
t h a n the syntax. In an example like
He stems casually --~ He's a casual thief
the original meaning of the source ~entence has been lost, and
casual has been cast loose as an a t t r i b u t i v e modifier of Animate
nouns; its historical origin determines one of its meanings, b u t is
irrelevant to the fact t h a t attribution selects one meaning (a casual
thie[) and predication another (The thie] is casual). The a t t r i b u t i v e
nl-~-cto~o 4"k,a'b .,~..~ ~n~o4- 4-~ rl .I . . . . .-1 . . . . "1~'~11.. .a. IL. . . . . y.._..~t -L2--.
~ ' . I J L IQL*..~r~O L I g O . L a & ~ ~ , x ~ g ~ L - u u e i £ v c =uvc, t,,,tuy, i.e. ttxe ==/,=r~

type, are precisely the ones where we need no special derivation,

for the situation is the same with non-agentive ,aouns; in
The young Augustine was a good t r e n c h e r m a n but a lousy saint
it would be rather difficult to regard saint as an agentive noun (to
handle it adverbially would call for *He sa~,~t-is lousily --~ He's a
lousy saint), yet He's a lousy skier is of exactly the same tylE.e. And
m a n y of the attributive phrases where we need some special
explanation, like large stockholder or clean /ighter, are o~.~es t h a t
(like home-loving girl or 90-mile-an-hour driver) suggest an act of
creation at a particular time and place. It does not appear to
m a t t e r what these historical origins were. A clever strategist is one
who plans strategy cleverly or who plans clever strategy or who is
clever at planning strategy. W h a t m a t t e r s for our present concern
is t h a t agentive nouns like all others select the meaning of attri-
butive adjectives on the basis of reference-modification. Since most
agentive norms embrace no more semantically t h a n the action of
the verb on which t h e y are based, an origin like
He grades strictly -+ He's a strict grader
is tempting as a S Y N T A C T I C way of accounting for all such noun
phrases. But once the agentive noun spreads its semantic range a
bit, this no longer wc,rks"
He teaches strictly -~ He's a strict teacher
 ADJECTIVES IN ENGLISH 31

A strict teacher m a y be one who merely keeps his pupils quiet. He is

strict in whatever sen:~e is relevant: to the noun.

8. A T T R I B U T I O N AND COMPOUNDING
There is a question whether any examples of the type medical
student industrial machinery, mgritime law, etc. are freely associated
adjective,~ and nouns rather than compotmds created in this way
instead of by drawing on some other resource such as noun +
noun. 90) If we assume that a given phrase is a compound, then it
follows that the adjective is inseparable and no predication will
necessarily relate to it. The Pentagon is a base of operations and it
is also military but it is riot a military base. Similarly
The tape is adhesive --~ adhesive tape
gives a plausible (and no doubt historically true) generation. But
it is no longer productive: we can say Scotch tape is adhesive, but
not *Scotzh tape is adhes;ve tape. Similarly a Shakespearian play
has lost the freedom of T.'~e ~klay is Shakesl~earian, which does not
have to mean 'by Shakespeare'. The typical case is the one where
the predication makes the wrong selection of c~tegory:
nervous system; This system is nervous
alimentary canal; *The canal is alimentary
ethical drugs; *These drugs are ethical
and so for in]lationary spiral, purple passage, and ca:~did camera:.
Some instances of this sort are like the word casual in the preceding
section: attributive position manifests a sense that is no longer
active:
a happy coincidence; *The coincidence is happy

s0) T h e r e seems t o be n o g o o d reason, for e x a m p l e , w h y t h e Civ,:l W a r

h a d n o u n + n o u n U n i o n Forces on o n e side a n d adj. + n o u n Con/~derate
_Forces o n t h e other, or a n y r e a s o n besides s p e e c h level w h y a m a n with a,
t i n h a t uses construction materials while o n e w i t h a cap a a d g o w n uses,
instructional materials - w o r d - f o r m a t i o n is a t r a n s f o r m a t i o n a l wilderness.
W e s a y a medical m a n for ' a d o c t o r ' b u t r, o t a dental m a n nor a surgical m a n
for ' a d e n t i s t ' a n d ' a s u r g e o n ' . W e keep a dental a p p o i n t m e m a n d a medical
a p p o i n t m e n t w i t h a dent~,st a n d a d o c t o r , b u t n o t a n *electrical a p p o i n t m e n t
w i t h a n electrician. T h e r e are legal m i n d s in t h e law b u t n o t *botanical m i n d s
a m o n g botanists.
32 DWIGHT BOLINGER

F o r some speakers/ortunate is p r o b a b l y going the same w a y :

a fortunate accident; ?The accident was fortunate
There seems to be a k i n d of polarization of H u m a n vs. N o n - h u m a n
Jn these and in
a brave sight; *The sight was brave; zl) The m a n was brave
a proud m o m e n t ; *The m o m e n t was proud
i t appears t h a t there is no w a y to d r a w a line between reference-
modification and compounds. A good case could be m a d e for
recognizing most instances of exclusi~rely a t t r i b u t i v e adjectives as
raw material for compounding. T h e y share w i t h obvious compounds
the inability to take the comparison,
cold cream, *colder cream
common noun, *commoner noun
military base, *more m ~ t a r y base
We can apply this test to narrow miss, which might seem to be
syntactically generated:
He fired. It was a narrow miss
He fired. It was a wide miss
These are both normal. B u t of
*That miss was narrow; you'll m a k e it n e x t time
T h a t miss was wide; you'll have ~c aim better n e x t time
ttle first gives me the same impression as
*That miss was near
based on a near miss. The comparative test applies:
*Tkat was a narrower miss
T h a t was a wider miss
and also the nominalization test:
T h a t miss was a narrow one
?That miss was a wide one
So it appears that narrow miss is a c o m p o u n d b y this reckoning,
while wide miss is net.

21) The fact that braw-loo/ci~g sight is no~'mal suggests t h a t brave was
originally predicative in both its s~nses.
 ADJECTIVES IN E N G L I S H 33

The argument is weakest with certain standardly classifying

adjectives which rather freely take predicative position. Adjectives
of nationality are the most noteworthy subclass:
These writers are American
The most famous paintings are Italian
This product is Canadian
And yet I note an inclination to nominalize, with a shift of stress
to the adjective, exactly as happens with compounds:
These are Am6rican writers
This is a Camldian product
The most famous paintings are the I't~lian ones.
And also a reluctance to use these adjlectives predicatively for
H u m a n subjects:
He's American; preferably, He's an .American
But where the adjective has the semantic feature of 'typical of
such people', it more readily goes with individual H u m a n referents:
Why does he behave that way ? - Because he's (so) Irish
In m y speech this is true of English, German, Scotch, Chinese, but
mot of A ustratian, American, Canadian, etc. Other similar nomi-
nalizations"
This is a m~nual sihift rather than This shift is manual
The: pass he threw was a l~itera! one rather than The pass he
threw was lateral
This is d6ntal equipment rather than This equipment is dental
The call I made was a s6cial one rather than The call I made
was social
(But, ip a slightly different sense, This call is purely social).
By thus broadening the definition of compound, we can relegate
most instances of the 'kind of + noun' generation (other *han the
good-bad ones), i.e. generation that ha.,, nothing to do with predi-
cation, to the level of word-formation, i.e., to the lexicon, leaving
referent-modification as a genuine syntactic transformational
source for attributive phrases that are not compounds, in addition
to others that do become compounds par'dy through contextual
34 DWIGHT BOLINGER

stereotyping in the course of time: ]reshman, wet nurse, easy chair,~2)

single ]ile, cold cream, silent partner, a likely story, a cool head. Com-
pounds in this sense would cover the spectrum of freedom, all
the way from a phrase like grammatical thinking (p~ 26 above),
which is almost free (though one would not have *grammatical
publishing, in spite of scienti]ic publishing), to one like heavenly host.

Harvard University

22) Easy chair lacks the manipulability of com~/ortable chair; *The chair is
easy, *an easier chair.