You are on page 1of 38

AL E XI S MANAS T E R - R AME R AND MI C HAE L B.

KAC
T HE CONCE P T OF P HRAS E S T RUCT URE
0. I NTRODUCTI ON
One of t he per enni al issues in t he st udy of human l anguage, debat ed
especi al l y in t he mi d fifties, and again in this decade, has been t he ad-
equacy of phr ase st r uct ur e as a model of synt ax. In this paper , we shall
ar gue t hat , f r om t he ver y begi nni ng, t her e has been conf usi on about
exact l y what t heor y was under discussion, so much so as t o cast doubt on
t he subst ant i ve nat ur e of this issue. Never t hel ess, we cont end t hat t her e
are real quest i ons concer ni ng t he power and t he equi val ence of vari ous
model s t hat have been r ef er r ed to as phrase structure grammars. In part i -
cul ar, this is so in r egar d to a l argel y t aci t concept i on of phr ase st r uct ur e
t hat has emer ged wi t hi n cur r ent synt act i c t heor y and whi ch appear s t o be
distinct f r om any of t he model s explicitly bear i ng this label. We will seek
t o capt ur e this el usi ve not i on by offeri ng an unf or mal i zed but r at her
det ai l ed decr i pt i on of a class of gr ammar s whi ch seems t o cor r es pond t o
it. We will show t hat this class of gr ammar s is unr est r i ct ed as t o weak
gener at i ve capaci t y but t hat it is sever el y rest ri ct ed in its ability to anal yze
cer t ai n ki nds of synactic pat t er ns. Whi l e t hese pat t er ns appear t o be r at her
c ommon in human l anguages, our goal will not be t o ar gue about t he
val i di t y of this par t i cul ar model of synt ax but r at her t o clarify j ust what it
is t hat needs t o be ar gued about .
1. T HE MANY ME ANI NGS OF ~ P HR AS E S T R UC T UR E '
The st or y of phr ase st r uct ur e begi ns with t he i nt r oduct i on, by Bl oomf i el d
(1914, 1933), of t he not i on t hat t he synt act i c st r uct ur e of a sent ence coul d
in most cases 1 be exhaust i vel y specified by anal yzi ng it i nt o t wo or mor e
i mmedi at e constituents (ICs) and t hen subj ect i ng each of t hese t o similar
analysis unt i l t he level of ultimate const i t uent s (t hat is, mor phemes ) is
1 It s houl d be not e d t ha t , st ri ct l y s pe a ki ng, Bl oomf i el d' s I C t he or y was not me a n t t o
cover all synt act i c p h e n o me n a . He deal t s e pa r a t e l y wi t h pr o- f or ms ( ' s ubs t i t ut i on' ) a nd wi t h
s e nt e nc e s as o p p o s e d t o cl aus es and phr a s e s , i . e. , " [ w] he n a f or m is s p o k e n al one ( t hat is,
not as a c ons t i t ue nt of a l ar ger f or m) [and] a ppe a r s in s ome sentence-O,pe" (1933, p. 169).
For s o me of t he ba c kgr ound of Bl oomf i el d' s i deas, s ee Per ci val (1966).
Linguistics and Philosophy 13: 325- 362, 1990.
1990 Kl uwer Academi c Publishers. Printed in the Netherlands.
326 ALEXIS MANASTER-RAMER AND MICHAEL B. KAC
r eached. Thi s i dea has had a var i et y of distinct f or mul at i ons and has
pr ove d except i onal l y dur abl e. I t was one of t he hal l mar ks of Ame r i c a n
st r uct ur al i sm, and has cont i nued t o ser ve as t he cor e of mos t synt act i c
model s t hat have e me r ge d si nce, such as t r ans f or mat i onal and uni fi cat i on
gr ammar s . Whi l e even t he necessi t y of this t ype of anal ysi s has f r om t i me
t o t i me been quest i oned (see, e. g. , Huds on 1976, 1984), it is t he quest i on
of its suffi ci ency t hat has ari sen as a maj or concer n in synt act i c t heor y.
The issue of whet her t he synt ax of huma n l anguages is a me na bl e to
descr i pt i on in t hese t er ms has on at l east t hr ee s epar at e occasi ons f ocused
t he at t ent i on of synt act i ci ans. Fi rst t he I C model was subj ect ed t o fi erce
cri t i ci sm by Choms ky (1956, 1957, 1961, 1962a, 1962b, and passi m) , who
ar gued f or a mor e compl ex f or mal i s m f or huma n l anguage synt ax, namel y,
t r ans f or mat i onal gr a mma r . Choms ky' s cri t i que was bas ed on a pur por t e d
f or mal i zat i on of I C anal ysi s in t er ms of st ri ng-rewri t i ng syst ems r ef er r ed
t o as phr as e st r uct ur e gr a mma r s ( PSG' s ) , and it consi st ed in a seri es of
ar gument s desi gned t o show t hat such gr a mma r s wer e i nher ent l y i ncapabl e
of suppor t i ng t heor et i cal l y adequat e anal yses of cer t ai n aspect s of huma n
l anguage synt ax. I n cer t ai n wor ks, t hese wer e t e r me d ' const i t uent st ruc-
t ur e gr a mma r s ' ( Choms ky 1962, 1963; Choms ky and Mi l l er 1963). Thi s
usage was f ol l owed by Post al (1964a) and Ha r ma n (1963), but Choms ky
(1965) and Ha r ma n (1966) r ever t ed t o ' phr as e st r uct ur e gr a mma r ' . 2
The r e have been ma ny r ej oi nder s t o Choms ky' s ar gument s , but onl y
on t wo occasi ons have full-scale explicit at t empt s been ma de t o show t hat
PSG coul d ser ve as an adequat e gener at i ve model of huma n l anguage
synt ax. The first was t he wor k of Ha r ma n (1963), whose "di scont i nuous-
const i t uent phr as e- s t r uct ur e g r a mma r wi t h subscri pt s and del et es " i ncor-
por at ed, as t he descr i pt i on i ndi cat es, sever al f or mal devi ces whi ch wer e
bar r ed by Choms ky' s def i ni t i on? Thi s def ense of phr as e st r uct ur e was
r ej ect ed by Choms ky (1965, 1966), who ar gued t hat t he augment at i ons
r esul t ed in a model t hat " has no mor e connect i on wi t h phr as e st r uct ur e
g r a mma r t han ant el opes have wi t h ant s" (1966, p. 41). Whi l e Ha r ma n ' s
second art i cl e i ncl udes a r es pons e t o Choms ky' s 1965 cri t i que (pp.
290-92), 4 t he debat e went no f ur t her at t he t i me, and Choms ky' s posi t i on
2 To add to the confusion, Harman (1963) used the term 'phrase structure grammar' to refer
not to Chomsky's formalism but to the IC model of the structuralists. This usage was
withdrawn in the 1966 article.
3 Harman's work was derived in part from that of Yngve (1960), where discontinuous
disconstituents had also been used.
4 We are grateful to Professor Harman for calling our attention to his 1966 paper. An
interesting point that we will discuss below is that in this paper he purports to argue for the
adequacy of PSG as defined by Chomsky, even though he reiterates his earlier position that
this definition did not do justice to the IC model.
THE CONCEPT OF PHRASE STRUCTURE 327
was wi del y accept ed. In fact , t he issue s eemed t o have been set t l ed until
t he case f or phr ase st r uct ur e was r e ope ne d with t he advent of gener al i zed
phr ase st r uct ur e gr ammar ( GPSG) . The necessi t y of t r ansf or mat i onal rules
was agai n cal l ed i nt o quest i on, and t he possibility of a pur el y phr ase
st r uct ur e amount of human l anguages was her al ded once mor e ( Gazdar ,
1981, 1982; Pul l um and Gazdar , 1982; Gazdar et al., 1985).
Thi s l at est devel opment pr ovi des us wi t h a fresh case of t he gener al
( and of t en difficult) pr obl em of deci di ng when t wo out war dl y di f f er ent
t heor et i cal model s are in fact equi val ent , so t hat ar gument s f or or against
one necessari l y appl y equal l y t o t he ot her . The hi st or y of t he phr ase
st r uct ur e cont r over s y is especi al l y enl i ght eni ng in this r egar d, and we feel
t hat its lessons shoul d i nf or m any debat e about such claims of equi val ence
whi ch may arise in t he f ut ur e. The i mpor t ance of t he phr ase st r uct ur e
case also st ems f r om t he i nher ent i nt er est t hat at t aches t o a st rong and
cont r over si al t heor et i cal claim: The r e is wi despr ead agr eement t hat a
convi nci ng demons t r at i on t hat human l anguage synt ax is - or is not -
amenabl e t o phr ase st r uct ur e analysis woul d have significant empi ri cal
consequences. Yet , at t he same t i me, Choms ky was surel y ri ght in his
admoni t i on t hat not hi ng of subst ance is t o be gai ned by redefi ni ng existing
t er mi nol ogy so as to obscur e t he subj ect of a debat e, and t he quest i on of
whet her in fact t he issues are t rul y subst ant i ve must be squar el y faced.
Most of t he discussion of phr ase st r uct ur e since Chomsky' s ori gi nal
cri t i que has f ocused on context-free grammar ( CFG) . Whi l e it has always
been cl ear t hat t her e exist non-CF phr ase st r uct ur e gr ammar s, t hese have
r eceded i nt o t he backgr ound. In part i cul ar, on each of t he t wo occasi ons
when t he phr ase st r uct ur e appr oach t o synt ax has been advocat ed, t he
specific model s pr opos ed have been equi val ent t o CFG' s in weak gener a-
tive capaci t y and have been explicitly adver t i sed as not at i onal vari ant s of
CFG' s. Si nce t he debat e began wi t h Chomsky' s cri t i que of t he I C model ,
however , it becomes crucial t o est abl i sh whet her this model was itself
mer el y a not at i onal var i ant of CFG, and so we need t o be cl ear, when
maki ng st at ement s about PSG, whet her t hese are about
(i)
or
(ii)
t he i mmedi at e const i t uent model of synt ax, assumi ng t hat this
was a single, coher ent model ,
t he class of cont ext - f r ee gr ammar s. 5
5 It should be noted that, while modern definitions of CFG follow Bar-Hillel et al. (1961)
in permitting null productions (or, more loosely, deletions), Chomsky has consistently (e.g.,
328 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
Ev e n i f t he t wo a r e e x t e n s i o n a l l y e q u i v a l e n t , t hi s c a n o n l y be d e c i d e d
b y c a r e f u l l y c o n t r a s t i n g t he o u t wa r d l y ( i n t e n s i o n a l l y ) d i f f e r e n t n o t a t i o n s .
Mo r e o v e r , t h e f i de l i t y o f t hi s f o r ma l i z a t i o n was c a l l e d i nt o q u e s t i o n b y
Ha r ma n , a n d i n d e e d , as n o t e d by Po s t a l ( 1964, Ch a p . VI ) , i t wa s t r ue
unt i l r a t h e r r e c e n t l y t ha t : " T h e ma j o r i t y o f l i ngui s t s a r e u n c o n v i n c e d t ha t
t he t h e o r y o f p h r a s e s t r u c t u r e c o r r e c t l y r e p r e s e n t s t he mo r p h o s y n t a c t i c
i d e a s wi t h wh i c h mo s t mo d e r n l i ngui s t s wo r k " . 6 I n d e e d , we wi l l al s o a r gue
t ha t ma n y a s p e c t s o f Bl o o mf i e l d i a n I C t h e o r y we r e a b s e n t f r o m Ch o ms k y ' s
g r a mma r s .
Be f o r e di s c us s i ng t h e s e i s s ues , h o we v e r , we n e e d t o p o i n t o u t t ha t
Ch o ms k y ' s a t t e mp t s at f o r ma l i z i n g I C a na l ys i s i n v o l v e d a b r o a d e r cl ass of
f o r ma l g r a mma r s t h a n t he CF G' s . Wh e n Ch o ms k y i n t r o d u c e d P S G' s , he
d e f i n e d t h e m so as t o a l l ow t he us e o f c o n t e x t ( 1956 [1965, pp. 111- 112] ) :
A phrase-structure grammar is defined by a finite vocabulary (alphabet) Vp, a finite set E
of initial strings in Vp, and a finite set F of rules of the form X~ Y, where X and Y are
strings in Vp. Each rule is interpreted as the instruction: rewrite X as Y. For reasons that
will appear directly, we require that in each such [E, F] grammar
(18) ~: E ~ . . . . . E,,
F: Xl --+ Y1
Xm----~ Ym.
Yi is formed from Xi by the replacement of a single symbol of X+ by some string. Neither
the replaced string nor the replacing string may be the identity element.
Thi s cl as s o f g r a mma r s , whi c h r e q u i r e s t h a t at e a c h s t e p i n t h e d e r i v a t i o n
a s i ngl e s y mb o l i n s o me c o n t e x t b e r e wr i t t e n as a n o n - n u l l s t r i ng, was
c a l l e d t y pe - 1 i n Ch o ms k y ( 1959) , b u t wa s s u b s e q u e n t l y r e n u mb e r e d t y p e - 2
i n Ch o ms k y ( 1963) . 7 I n a d d i t i o n , Ch o ms k y ha s c a l l e d t h e s e g r a mma r s
c o n t e x t - r e s t r i c t e d ( 1961) o r c o n t e x t - s e n s i t i v e (1963 a n d e v e r s i nce) .
Ho we v e r , t h e r e ha s a p p a r e n t l y b e e n s o me c o n f u s i o n a b o u t t hi s p o i n t ,
as e v i d e n c e d b y t h e f act t h a t t he t e r m p h r a s e s t r u c t u r e g r a m m a r is o f t e n
1956, 1959) defined CFG' s, like CSG's, so as to exclude this possibility. Also, regular
expressions have never been allowed by Chomsky to appear on the right-hand sides of CF
productions.
6 Postal (p. 72) specifically cites Pike (1954-60, III, p. 36) and Hockett (1961, p. 230).
7 Of course, in the 1959 numbering scheme, type-2 referred to context-free grammars, which
became type-4 in 1963. The 1961 type-1 grammars are a proper superset of the CSG's,
defined by the requirement that the right-hand side of each rule not be shorter than the left-
hand side. There is no difference in weak generative capacity between these two classes,
and the distinction between them will not play any role in the subsequent discussion here,
though it behooves us to note that many theoretical computer scientists use the term con-
text-sensitive grammar to refer to the broader class (e.g., Hopcroft and Ullman, 1979, p.
223).
T HE C ONC E P T OF P HR AS E S T R UC T UR E 329
- in theoretical computer science, almost always (e.g., Harrison, 1978, p.
13; Hopcroft and Ullman, 1979, p. 220; Berwick and Weinberg, 1982, p.
168, n. 3) - used to denote type-0 (unrestricted) grammars, which generate
all the recursively enumerable languages, a proper superset of the context-
sensitive languages, s The reason seems to be that, while Chomsky (1956)
clearly only dealt with CSG's, his later works sometimes offer informal
descriptions of PSG's that leave out crucial conditions (such as the con-
straint against null replacement, e.g., 1957, p. 29) or would present the
formal definition of a ' grammar' (i.e., a type-0 grammar) either immedi-
ately after or even in the midst of an extended prose discussion of PSG's,
without emphasizing the distinction (e.g., 1959 [1965, p. 128-29]). Since,
moreover, type-0 grammars were not even considered in the 1956 paper,
it might have seemed that, in his later works, Chomsky was in fact broad-
ening the scope of the term phrase structure grammar to include them.
For example, (e.g., 1959 [1965, p. 128]):
A phr as e st r uct ur e gr a mma r consi st s of a finite set of ' r ewr i t i ng r ul es' of t he f or m &---~ ~,
wher e 05 and ~ ar e st ri ngs of symbol s. It cont ai ns a speci al ' i ni t i al ' symbol S ( st andi ng f or
' s e nt e nc e ' ) and a bounda r y symbol # i ndi cat i ng t he begi nni ng and end of s ent ences .
Since nothing is said about the relative lengths of 4~ and t), this appears
to refer to type-0 grammars. In subsequent passages additional restrictions
on the form of rules are introduced, which do refer to CSG (pp. 128-29):
If appr opr i at e r est r i ct i ons ar e pl aced on t he f or m of t he rul es 0 5 ~ 0 (in par t i cul ar , t he
condi t i on t hat qJ di ffer f r om 49 by r e pl a c e me nt of a single symbol of q5 by a non-nul l st ri ng),
it will always be possi bl e t o associ at e wi t h a der i vat i on a l abel ed t r ee in t he same way.
But on a first reading, it may not be clear that Chomsky is still defining
PSG' s in general rather than a special subclass of these. Only a careful
reading rules this out, as does the explicit definition of constituent-struc-
ture grammars in Chomsky (1961, p. 9), which reiterates the constraints
stated in the 1956 paper:
Suppos e t hat each synt act i c rul e 05~ 0 meet s t he addi t i onal condi t i on t hat t her e is a single
symbol A and a non- nul l st ri ng w such t hat & = xI Ax2 and ~ = XlW/tv2 . . . . A set of rul es
meet i ng t hi s condi t i on I will call a constituent structure grammar. If in each rul e 05-+ 0, 05
is a si ngl e symbol , t he gr a mma r ( and each rul e) will be cal l ed context-free; ot her wi s e,
context-restricted.
In light of all this, it appears that Chomsky consistently understood PSG's
8 Some c omput e r sci ent i st s, however , like t o use t he t er m type-O t o r ef er t o s ome pr ope r
subset of unr es t r i ct ed gr ammar s . Thus, Har r i s on (1978, p. 17) appl i es it t o t he class of
gr ammar s in whi ch t he l ef t - hand si des of pr oduct i ons may not cont ai n t er mi nal symbol s. But
t hes e gr ammar s still gener at e all t he r . e. l anguages.
330 ALEXI S MANASTER- RAMER AND MI CHAEL B, KAC
to be formal l y defi ned as CSG' s i n t he nar r ow sense, i . e. , gr ammar s whi ch
repl ace a single symbol by a non- nul l st ri ng i n some (possi bl y nul l ) cont ext ,
at each st ep i n t he deri vat i on. Thi s gives us a t hi rd not i on of PSG:
(iii) cont ext -sensi t i ve gr ammar .
However , al t hough strictly speaki ng non- CSG' s were not consi der ed
PSG' s, we will show t hat Chomsky' s ar gument s agai nst phrase st ruct ure
as a model of huma n l anguage synt ax appl y j ust as much to t ype-0 gram-
mars as t hey do to CSG' s or CFG' s .
The st ory does not end t here, for i n s ubs equent work Chomsky hi nt ed
at a class of gr ammar s mor e rest ri ct ed t han CSG' s but mor e gener al t han
CFG' s as t he appr opr i at e model of I C anal ysi s. These were l abel ed t y p e - 3
g r a m m a r s , because t hey were defi ned by a Condi t i on 3 whi ch di st i ngui shed
t hem f r om t y p e - 2 (i . e. , cont ext -sensi t i ve) gr ammar s (1963, p. 366): 9
Condition 3 G' is a type 2 grammar containing no rule x1Ax2x1~oX2, where ~o is a single
nonterminal symbol (i.e., v E VN).
The mot i vat i on for t hese gr ammar s was to per mi t cont ext ual rest ri ct i ons
wi t hout al l owi ng, as i n t he gener al class of CSG' s, t he possi bi l i t y of
per mut at i on. The fact t hat t hese gr ammar s are mor e powerful t han CFG' s
was t hus part of t hei r appeal (i bi d):
Only one nontrivial property of type 3 grammars in known, namely, that stated in Theorem
14. This class of grammars merits further study, however. It seems that Condition 3 provides
a reasonably adequate formalization of the set of linguistic notions involved in the richest
varieties of immediate constituent analysis?
I n fact, i n t he same paper , Chomsky ar gued t hat even this non- cont ext
free model was still t oo rest ri ct ed for huma n l anguage (i bi d):
Condition 3, as it stands, is too strong to be met by actual grammars of natural languages,
but it can be revised, without affecting generative capacity, to be perhaps not unreasonable
for the construction of grammars of language-like systems. Suppose that we allow the
grammar G to contain a rule x1Ax2---~Xao~x2 only when a is either terminal (as in Condition
3) or when a dominates only a finite number of strings in the full set of derivations (and P-
markers) constructible from G. This essentially amounts to the requirement that if a category
is divided into subcategories these subcategories are not phrase types but word or morpheme
classes. To the extent that systems of the kind we are now discussing are at all useful for
9 The reader should beware of any confusion due to the fact that in the more familiar 1959
numbering system, type-3 refers to regular grammars and type-2 to CFG's.
l0 Chomsky's Theorem 14 holds that the language L = {ac"f2+4"d~b, n >1 0}, which is not
context-free, is a type-3 language. From this it follows that the proper containment of
grammar types extends to classes of languages, that is, that the type-3 languages are a proper
superset of the CFL's.
THE CONCEPT OF PHRASE STRUCTURE 331
grammatical description, it seems likely that the particular subclass meeting this condition
will in fact suffice.
To do justice to the historical facts, then, we must take into account
some further possible notion of PSG:
(iv) some proper superset of CFG' s such as Chomsky's 'type-3'
grammars.
These particular ideas of Chomsky's appear to have been generally forgot-
ten, but it is uncanny that in recent years there have been a number of
proposals to the effect that human languages are 'mildly context-sensitive',
i.e., not contained within the class of CFL' s but within some class which
is larger than the CFL' s and itself properly contained within the CSL's
(Joshi, 1985; Gazdar and Pullum, 1985; Gazdar, 1985).
As we have seen, Chomsky's notion of phrase (constituent) structure
was supposed to correspond to the theory of immediate constituents.
Indeed, in his early works (1956 [1965, p. 111], 1957, p. 34, 1959 [1965,
p. 132]), Chomsky refers to IC analysis by name. In later publications
(1965, 1966), the criticism is directed more generally at 'structuralist' or
'taxonomic' grammars, but throughout it is clear that PSG was supposed
to be the formalization of a pre-existing theoretical concept. We have also
seen that, while Chomsky tried to restrict the formal model of PSG in
certain ways, it always remained non-context-free. Given this, it might
seem mysterious that, as noted above, the debates about the adequacy of
phrase structure have focused on CFG. The mystery is solved when we
examine some other of Chomsky's writings, especially the linguistic (as
opposed to mathematical) ones, for in several of these the claim is made
that actual IC systems were purely context-free.
We first find this position in a paper read as early as 1958 but published
some years later, where Chomsky (1962b) presents context-sensitive gram-
mar as a formalization of Harris's (1951) morpheme-to-utterance formulae
but argues that actually Harris's model was context-free. The context-
sensitivity is offered as a modification of Harris's original system, but one
made "in accordance with his intentions" (p. 129). Here Chomsky seems
to be alluding to Harris's observation (1946, p. 182) that "It]he great bulk
of selection features, especially those that distinguish between individual
morphemes, cannot be expressed except by very unwieldy formulae". The
context-sensitive rules would then be used to capture the selectional facts
missed by Harris's context-free formulae.
However, we know that Chomsky regarded Harris's model as having
other, deeper limitations, which could not be overcome except by the
332 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
i nt roduct i on of t ransformat i ons. Appar ent l y, t hen, in or der to focus the
discussion on these more serious defect s, he woul d direct his criticism at
a model which had t he needed context-sensitive feat ures built in. Thus,
t he (context-sensitive) PSG' s woul d not strictly speaki ng formal i ze the
mor pheme- t o- ut t er ance model . Rat her , the most obvious short comi ng of
t he l at t er woul d be pat ched up by the use of context-sensitive rules to
capt ure selectional restrictions, and the discussion could t hen be focused
on what Chomsky percei ved to be the real ar gument s for t he superi ori t y
of t ransformat i onal gr ammar . However , none of this was made clear at
the time.
At t he same t i me, Chomsky r epeat edl y describes his PSG' s as represent -
ing not just t he mor pheme- t o- ut t er ance syst em but also linguistic analysis
" on t he IC l evel " (p. 129; see also p. 129, no. 11, p. 130, p. 131), wi t hout
specifying in det ai l t he rel at i on of t he IC and the mor pheme- t o- ut t er ance
approach. It woul d seem t hat he was accepting Harri s' s (1951, p. 278)
assertion t hat t he mor pheme- t o- ut t er ance and IC model s di ffered onl y
in di rect i on (bot t om-up vs. t op-down). Thus, Chomsky was apparent l y
claiming at this t i me t hat bot h t he mor pheme- t o- ut t er ance and IC t heori es
were cont ext -free in l et t er, but context-sensitive in i nt ent . Essent i al l y the
same position was t aken by Chomsky (1961, p. 9, no. 8):
Immediate constituent analysis as developed within linguistics, particularly in the form given
to this theory by Z. S. Harris, Methods in Structural Linguistics, (Chicago, Univ. of Chicago
Press, 1951), Chap. 16, suggests a form of grammar similar to what is here called context-
free constituent structure grammar.
Her e, t oo, he adds the qualification t hat "cont ext -rest ri ct ed rules are
unavoi dabl e, in practice, in grammat i cal descri pt i on" ( i bi d. ) .
The same view is rei t erat ed in t he first part of Chomsky (1962a), al-
t hough here his criticism is di rect ed at what he calls t he ' t axonomi c model '
of generat i ve grammar. This model is context-sensitive, i nasmuch as ' The
syntactic component consists of an unor der ed set of rewriting rules, each
of which states the member shi p of some phrase cat egory or format i ve
cat egory in some cont ext " (p. 510). The t axonomi c model is present ed as
"a direct out growt h of moder n st ruct ural linguistics" and " no more t han
an at t empt to f or mul at e a generat i ve gr ammar which is in the spirit of
moder n procedural and descriptive appr oaches". Specifically, it is again
said to const i t ut e an ext ensi on of Harri s' s mor pheme- t o- ut t er ance system.
However , for the first t i me, Chomsky appears to make a distinction
bet ween t he mor pheme- t o- ut t er ance t heor y, which he cont i nues to t ake
as context-sensitive in spirit, from IC model s proper, which he apparent l y
considers to be cont ext -free t o u t c o u r t (p. 557, fn. 3):
THE CONCEPT OF PHRASE STRUCTURE 333
On the syntactic level, the taxonomic model is a generalization from Harris' morpheme-to-
utterance statements, which constitute the nearest approach to an explicit generative gram-
mar on this level. Furthermore, most modern work in syntax is actually more adequately
formalized in terms of rewriting rules with null context (in particular, this is true of Pike's
tagmemics, as of most work in IC analysis).
Unf or t unat el y, Choms ky did not st at e this di st i nct i on cl earl y, or say any-
t hi ng mor e specific about I C t heor y. I n par t i cul ar , he di d not ci t e any
specific names of schol ars or titles of publ i cat i ons associ at ed wi t h I C
anal ysi s. To be sure, in a l at er r evi si on of t he s ame wor k, Choms ky (1964,
p. 916, n. 4) ampl i fi es this s t at ement wi t h a r ef er ence t o Post al ' s (1964a)
det ai l ed cri t i que of six nont r ans f or mat i onal model s of synt ax.
Howe ve r , Post al ' s account cont r adi ct s Choms ky' s cl ai ms, and ma ke s
t he pr i ma faci e case t hat bot h I C and mor phe me - t o- ut t e r a nc e anal ysi s di d
i nvol ve cont ext - sensi t i vi t y. Of t he six st r uct ur al i st appr oaches he exam-
i ned, Post al cl ai med t hat onl y t hr ee wer e cont ext - f r ee, and t hese wer e all
model s devel oped at r oughl y t he s ame t i me as or l at er t han t r ansf or -
mat i onal gr a mma r : t he t agmemi cs of Pi ke (1954-60), t he st at i fi cat i onal
g r a mma r of La mb (1962), and Hoc ke t t ' s (1961) const r uct i onal gr a mma r .
The r emai ni ng pos t - t r ans f or mat i onal model cons i der ed was Har r i s ' s st ri ng
anal ysi s (1962), whi ch Post al bel i eved t o be cont ext - sensi t i ve.
Of t heor i es ear l i er t han t r ans f or mat i onal gr a mma r , Post al di scusses
Har r i s ' s mor phe me - t o- ut t e r a nc e model and t he I C model pr oper . For t he
l at t er, Post al does not r ef er t o t he ori gi nal t heor et i cal and descr i pt i ve
wor k of Bl oomf i el d, but onl y t o Bl och (1946), Wel l s (1947), and Hoc ke t t
(1954), 11 but he still concl udes (p. 79) t hat t he I C model - like t he
mor phe me - t o- ut t e r a nc e one - was cont ext - sensi t i ve and cont r ast s it on
this ver y poi nt wi t h t he t hr ee l at er , cont ext - f r ee model s (ibid). Mor e ove r ,
he cites (p. 19) an exampl e of what he t akes t o be cont ext - sensi t i vi t y
(see bel ow) f r om Bl och. 22 Fi nal l y, wri t i ng about t he cont ext - sensi t i vi t y of
Har r i s ' s mor phe me - t o- ut t e r a nc e f or mul ae ( pp. 25-26), Post al cont r adi ct s
Choms ky' s posi t i on t hat cont ext - sensi t i vi t y was onl y i mpl i ci t in this ki nd
of wor k by citing t he expl i ct use of it in Har r i s ' s wri t i ngs. I f Post al was
ri ght , t hen cont ext - sensi t i vi t y was par t and par cel of bot h I C and mor -
phe me - t o- ut t e r a nc e syst ems of synt act i c anal ysi s.
I n poi nt of fact , t her e is no doubt t hat Har r i s ma de ext ensi ve use of
n Postal refers to Hockett's model as the "item and arrangement system", but Hockett is
quite clear that he is talking about IC analysis.
t2 As we mention below, Postal apparently mistook a case where Bloch (tacitly) refers to
cross-classification, for context-sensitivity. This is apparently the only case where Postal
claims to find context-sensitivity in IC work.
334 AL E XI S MANAS T E R - R AME R AND MI C HAE L B. KAC
context-sensitivity as a way of controling the substitutions allowed by his
morpheme-to-utterance formulae (e.g., 1946, pp. 166-67):
Equa t i ons will be u s e d t o i ndi cat e subst i t ut abi l i t y. BC=A will me a n t hat t he s e que nc e
cons i s t i ng of a mo r p h e me of cl ass B f ol l owed by a mo r p h e me of cl ass C can be s ubs t i t ut e d
f or a si ngl e mo r p h e me of cl ass A . . . . Wh e n we wa nt t o s ay t hat A s ubs t i t ut e s f or B onl y if
C f ol l ows, we shal l wr i t e AC = BC.
This means that contextual restrictions were known before the advent of
generative grammar. To be sure, neither Bloomfield nor Bloch seem to
have made much use of them. Perhaps the clearest example would be
Bloch's analysis of the zero variant of the copula in Japanese (1946, p.
213):
Bef or e t he par t i cl e ka . . . . t he non- pa s t i ndi cat i ve of t he copul a, da, is r epl aced by a zer o
a l t e r na nt ( i . e. , dr ops out ) .
Thus, on Postal's account - and he appears to have been right - the
use of context-sensitivity in PSG's would have been nothing more than
simple accuracy in formalizing either the morpheme-to-utterance or the
IC model. 13
In subsequent work of Chomsky's, we find no direct reference to IC
analysis at all, but we do find the sweeping generalization that "it has also
been shown that almost all of the nontransformational syntactic theories
that have been developed within modern linguistics, pure or applied, fall
within this framework" (1965, p. 67) and specifically, that "every variety
of syntactic theory that falls within the general range of taxonomic syntax
seems to me to be formalizable (insofar as it is clear) within the framework
of phrase structure grammar (in fact, with rare exceptions, its context-
free variety)" (1968, p. 39).
This is the position that seems to have been tacitly accepted by much
of the field, although it does not correspond to the explicit definitions of
phrase structure stated by Chomsky. As a result, CFG' s alone have figured
in subsequent debates over the adequacy of phrase structure models.
Thus, both Harman and the generalized phrase structure grammarians
explicitly undertook to defend the context-free model. More generally, it
13 Mc Ca wl e y (1968) has a r gue d t ha t t he us e of cont ext in wha t he me r e l y des cr i bes as
" mo s t Ame r i c a n st r uct ur al i st g r a mma r s of t he 1940' s" a mo u n t e d t o cont ext - s ens i t i ve node
admi ssi bi l i t y condi t i ons , whi ch ha ve t he we a k gener at i ve capaci t y of CF G' s ( Pet er s a nd
Ri t chi e, 1969). Thi s mi ght a ppe a r t o r e nde r t he i ssue moot , unt i l one r eal i zes t ha t t he
equi val ence r esul t s pe a ks onl y t o t he cl ass of l a ngua ge s t hat can be ge ne r a t e d. It s houl d be
a ppa r e nt t hat cont ext - s ens i t i ve node admi ssi bi l i t y s ys t e ms can r e pr e s e nt sel ect i onal and
a g r e e me n t p h e n o me n a i n ways i mpos s i bl e in CFG, so t he i ssue of wha t t he s t r uct ur al i s t
mode l s wer e capabl e of woul d still r e ma i n.
T HE C ONC E P T OF P HR AS E S T R UC T UR E 335
has become part of the folklore of syntax that that is what the phrase
structure controversy is all about, as though the arguments against phrase
structure were specifically concerned with the limitations of c o n t e x t - f r e e
models (e.g., Savitch et al . , 1987, pp. vii-viii; Pullum and Gazdar, 1982,
pp. 472-475, and the references cited there, Bresnan et al . , 1982, p. 613).
This is turn would seem to be the reason why questions of weak generative
capacity have so often been seen as bearing on the issue of the adequacy
of phrase structure models for human language syntax. However, as we
will show in the next section, the arguments raised by Chomsky against
PSG are by no means so restricted. They apply with equal force to context-
sensitive, to type-l, and even to type-0 (unrestricted) grammars. This
means, of course, that weak generative capacity cannot be the issue, since
no formal grammar has greater weak generative capacity than type-0
grammars.
2. ON T HE L I MI T AT I ONS OF P HR AS E S T R UC T UR E
Given what has been said about the history of the definition of PSG, it
should be clear that the classic arguments against this model were directed
specifically against context-sensitive grammars. The confusion about this
point has obscured the full force of these arguments, for Chomsky's case
against phrase structure, if accepted, establishes the inadequacy of much
more than just CFG' s. In fact, although this was apparently never stated
at the time, almost all of these arguments apply with equal force to type-
0 grammars. In particular, this would seem to be the case for arguments
involving unbounded branching (constituents with arbitrarily many im-
mediate subconstituents), discontinuous constituents, separate provision
for linear order as opposed to constituency, and cross-classification.
Unbounded branching is involved, on most analyses, in coordinate
structures (such as J o h n a n d Ma r y a n d Sal l y a n d B u r t . . . ) . Chomsky and
Miller (1968, p. 298) took the inability of PSG's to derive such structures
as a telling argument against them. It should be clear that a grammar that
contains rules of the form
NP---~ NP(CONJ NP)*
would account for the phenomena in question at least as well as a transfor-
mational system. Moreover, we now know that allowing such rules in a
CFG does not increase its weak generative capacity. But such rules were
not allowed by the letter of Chomsky's definition of either CSG's or type-
0 grammars, any more than by his definition of CFG' s.
Discontinuity is involved in the well-known arguments for the transfor-
336 ALEXIS MANASTER-RAMER AND MICHAEL B. KAC
mat i onal analysis of Engl i sh auxi l i ary verbs and t he suffixes whi ch t hey
demand on cooccur r i ng ver b stems. The facts as descr i bed by Choms ky
(1957) are t hat cert ai n funct i ons in t he ver bal syst em ( per f ect , progressi ve,
and passive) are char act er i zed by compl exes consisting of an i ndependent
auxi l i ary ver b and a suffix at t ached t o an adj acent ver b st em (viz. have V-
en, be V-i ng, and be V-en). Choms ky used cont ext - f r ee base rul es t o
deri ve have-en, be-ing, and be-en as cont i nuous el ement s and t hen l et a
t r ansf or mat i on t r anspor t each suffix t o t he end of t he appr opr i at e v e r b
st em, but t he same facts coul d be handl ed by describing have. . - en,
be. . - i ng, and be. . - en, as di scont i nuous const i t uent s. As Choms ky hi msel f
not es, "i n t he auxi l i ary ver b phr ase we real l y have di scont i nuous el ement s
- e. g. , . . . t he el ement s have. . en and b e . . i n g " (1957, p. 41).
As anot her exampl e, consi der Russi an yes/ no quest i ons, which di ffer
f r om st at ement s, in t hat t he mor phe me li appear s inside t he clause, e. g. ,
Prigol-li Ivan? ' Di d I van come?' . The r e are formal i sms whi ch woul d allow
us t o anal yze this sent ence as bei ng compos ed of t he t wo const i t uent s li
and prigol Ivan (which by itself is equal in f or m to t he st at ement ' Ivan
came' ) . Such anal yses can be st at ed in a model of gr ammar which allows
di scont i nuous const i t uent s, such as t he weakl y cont ext - f r ee gr ammar s of
Yngve (1960) and Ha r ma n (1963), but not in any t ype of Chomsky' s
grammars. Whe t he r t ype-0, cont ext -sensi t i ve, or cont ext - f r ee, t he l at t er
oper at e on strings, and rest ri ct rewri t i ng t o t he r epl acement of a cont i nu-
ous string of gr ammar symbol s by anot her such string. 14
Mor e general l y, many of t he ar gument s f or t r ansf or mat i ons have hi nged
on dependenci es bet ween el ement s whi ch are not sister const i t uent s. I f
we consi der a l anguage wher e quest i ons di ffer f r om st at ement s in wor d
or der , like Ge r ma n Ist Hans gekommen? ' Di d Hans come?' (vs. Hans ist
gekommen ' Hans came' ) , we see t hat di scont i nuous const i t uent s are not
enough. In such cases, it is simply t he di f f er ence in wor d or der whi ch
distinguishes quest i ons f r om st at ement s, and t her e are formal i sms whi ch
can st at e this fact di rect l y, by handl i ng const i t uency and l i near or der
separ at el y (e. g. , t he GPSG of Gazdar et al., 1985). However , like CFG' s,
bot h CSG' s and t ype-0 gr ammar s oper at e on t ot al l y or der ed st ruct ures,
and cannot handl e such phe nome na in t he same way.
Cross-classification is anot her phe nome non which figures pr omi nent l y
14 By comparison, consider languages in which yes/no questions differ from statements by
containing a marker which is clause-initial (e.g., Polish czy) or clause-final (e.g., Japanese
ka). In such cases, even a simple CFG could analyze the question marker as one IC and the
rest of the clause as another.
THE CONCEPT OF PHRASE STRUCTURE 337
in arguments for transformations. ~5 Two such arguments (Chomsky, 1956)
have become especially famous. One involves nominalizations like the
shooting of the hunters in relation to clauses such as The hunters shoot
and They shoot the hunters. While Chomsky argued that the ambiguity of
the nominal constructions had to be accounted for by assuming that they
derive from deep structures which look more or less like the verbal con-
structions, it should be apparent that an analysis involving complex sym-
bols is also possible. Since Chomsky was concerned with expressing the
fact that the hunters can be either subject or object, we could introduce
corresponding features and make this phrase be either of the complex
category [NP, subject] or [NP, object]. Alternatively, we could also refer
to sentences like The hunters are shot instead of They shoot the hunters
and use features to assign shooting to two different categories such as [N,
active] and [N, passive]. If rules of the grammar are allowed to refer to
such features, then it becomes possible to handle the verbal and nominal
constructions as related without recourse to transformations.
The other argument of Chomsky' s involves the relation between active
and passive clausal constructions themselves. Again, the dependencies
here involve differing word orders and cooccurrence restrictions on non-
sister constituents (the be..en auxiliary element and transitive verbs, for
example). The use of complex categories with features such as [active]
and [passive] and possibly also [agent] and [patient] would allow the same
general statements to be made as was possible under a transformational
analysis. 16 It becomes clear, incidentally, that what is responsible for the
apparent problems in analyzing the active/passive relation is the particular
word order and cooccurrence facts, rather than some inherent propert y
of voice as a grammatical notion. If there were to exist a language in
which actives and passives differed only the way that statements and yes/no
questions do in Polish or Japanese, then a phrase-structural analysis would
15 Moreover, many of the phenomena that might seem to be amenable to analysis in terms
of discontinuous constituency or of separate word order rules may actually involve cross-
classification. For example, many would argue that in the English auxiliary system it is
desirable to treat the V-en and V-ing sequences as word-level constituents. This is, of course,
precluded by the analysis in terms of discontinuous elements. However, the use of a suitable
feature system of categories would allow such an analysis.
a6 The limiting case would be the use of complex categories to encode directly the steps
in a transformational derivation, with features like [underlying], [derived-by-the-Passive-
transformation], etc. Of course, to get the full weak generative capacity of transformational
grammars, an infinite number of complex symbols would be required.
338 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
be perfectly possible.17 Such a language may not be attested, and perhaps
it is not even typologically possible, but the fact remains that the difficulties
with handling the active/passive relation in a PSG had to do with its
inability to handle certain kinds of the word order and cooccurrence facts.
And, once again, we find that cross-classification can be expressed in
weakly CF models, such that of Harman (1963) or Gazdar et al. (1985),
but not in CFG, CSG, or type-0 grammar, all of which empl oy unanalyzed
nonterminal symbols.
To summarize, type-0 grammars fare no bet t er than CFG' s or CSG' s on
Chomsky' s arguments, since they do not allow any of the four grammatical
devices suggested above: discontinuous constituents, separation of linear
order from constituency, complex nonterminal symbols (cross-classific-
ation), or unbounded branching. 18 As a result, if we assume that any of
these phenomena must be directly represent ed in grammars of human
languages, then these would have to be more powerful in some sense than
type-0 grammars. Now, the Church-Turing thesis, together with the fact
type-0 grammars enumerat e t he same class of languages that Turing ma-
chines do, implies that this is precisely the class of languages that can be
enumerat ed by any sort of computation whatsoever.
Hence, Chomsky' s arguments against PSG must be logically indepen-
dent of considerations of weak generative capacity. This is a point of
some importance, since, if correct, it shows that we cannot take literally
Chomsky' s goal of finding a place for the grammars of human languages
in the hierarchy which has come to be known by his name (1959 [1965,
p. 127]):
Given such a classification of special kinds of Turing machines, 19 the main problem of
immediate relevance to the theory of language is that of determining where in the hierarchy
of devices tile grammars of natural languages lie. It would, for example, be extremely
17 This would be a hypothetical language in which the agents and patients occupy the same
positions in active and passive sentences, but the latter are marked by a special sentence-
initial or sentence-final particle. Thus, the hypothetical passive of John saw Mary might be
John saw Mary pa (where pa is the hypothetical passive morpheme).
is Early arguments for transformations also sometimes referred to deletions. However, the
situation here is different than in the case of the other phenomena considered, because these
are allowed by Chomsky's definition of type-0 grammars. Any argument for the necessity of
deletions would, to be sure, still count against CSG's, but it would not apply to type-0
grammars.
19 This refers to the hierarchy of grammars, which Chomsky is presenting as special cases
of the Turing machine formalism.
T HE C ONC E P T OF P HR AS E S T R UC T UR E 339
interesting to know whether it is in principle possible to construct a phrase structure gram-
mar 2 for English (even though there is good motivation of other kinds for not doing so).
If even type-0 grammars are inadequate as models of human language,
then the correct class of models, whatever they may be, will not be
anywhere in this hierarchy. Chomsky must have meant not the hierarchy
of devices, but the corresponding hierarchy of languages, although even
then it might well be that human languages as a class are incomparable
to the types of languages defined by his hierarchy.
It is worth recalling in this regard that the case against PSG argued by
Chomsky in the late fifties made no reference to weak generative capacity.
The kinds of arguments we have talked about in connection with un-
bounded branching, discontinuity, cross-classification, and linear order
never claimed that a certain language could not be weakly generated by
a PSG. Indeed, while argument s directed specifically against CFG came
to rely on claims of weak inadequacy during the sixties, the quotation
above shows that as late as 1959 Chomsky was still considering the possibil-
ity that human languages are context-sensitive. The reasons for rejecting
PSG' s as a model of human language had to do with theoretical motiva-
tions which find partial formal expression in the notion of strong generative
capacity but on the whole have remained outside the scope of formal
language theory.
As a result, the divisive and much-debat ed issue of whether human
languages are context-free or not has little to do with the question of the
adequacy of PSG' s. Hence, questions of what weak generative capacity is
required to model human languages have no bearing on the adequacy of
phrase structure models for such languages.
3. S OME NE GL E C T E D AS P E C T S OF I C ANAL YS I S
We return now to the question of the extent to which the phrase structure
debate has been confused from the very beginning by the fact that Chom-
sky' s model of PSG failed to do justice to the structuralist theories and
practices which it was supposed to formalize. We have seen that there
20 I.e., CSG. Postal (1964a, p. 76, 1964b) would later claim that Mohawk was probably not
a CSL, though he never advanced an explicit argument for this. All this indicates that there
may have been some unclarity about the full extent of the weak generative capacity of
CSG's, for a time. Cf. Chomsky's remarks a few years later: "Although type 1 grammars
generate only recursive sets, there is a certain sense in which they come close to generating
arbitrary recursively enumerable sets" (1963, p. 361), or the classic formulation by Hopcroft
and Ullman: "'Almost any language one can think of is context-sensitive..." (1979, p. 224).
340 AL E XI S MANAS T E R - R AME R AND MI C HAE L B. KAC
has been a persistent discrepancy between Chomsky's claims that the
approaches he was formalizing were really context-free and his practice
of allowing context-sensitivity in the formalism. However, context-sensitiv-
ity is in fact a clumsy way of achieving the descriptive power which
Bloomfield had without it, and the real shortcoming of Chomsky's pro-
posed formalization of both IC and morpheme-to-utterance work will turn
out not to be the muddle over context-sensitivity but the fact that he
simply left out numerous other devices which the structuralists used to
account for complex grammatical phenomena. These included discontinu-
ous constituency, separate treatment of linear order and constituency,
cross-classification, unbounded branching and null (zero) elements.
Unbounded Branching
When Chomsky was marshalling his arguments against PSG, one of the
most telling was the inability of such grammars to provide for unbounded
branching (Chomsky and Miller, 1963, p. 298). Moreover, Harman ac,
cepted this aspect of Chomsky's model of phrase structure and did not
provide any device for unbounded branching, a point which Chomsky
(1965, p. 196) in fact raised against him. As is well known today, however,
the augmentation of any class of PSG's with regular expressions formed
with the Kleene star operator (which denotes concatenation closure) al-
lows for unbounded branching to be induced, and these devices have been
included in both transformational and nontransformational grammars (in-
cluding GPSG).
Yet IC analyses of the structuralist period routinely allowed construc-
tions with an unbounded number of immediate constituents. For example,
Bloch (1946, p. 207) takes a Japanese sentence to consist of two IC's, the
first of which is described as consisting of "as many constituents as there
are non-final [i.e., subordinate] clauses", and likewise treats Japanese
clauses as composed of a predicate IC preceded by an IC comprising "as
many constituents as there are clause attributes". Since the number of
subordinate clauses or clause attributes is unlimited, the first IC in each
case may itself consist of an unbounded number of IC's. Perhaps even
more telling is the way that Bloomfield analyzed coordinate constructions,
which were precisely the forms that Chomsky wanted unbounded branch-
ing for. Bloomfield's description unmistakably implies the possibility of
any number of IC's, for he treats coordinate constructions as having as
many heads as they have conjuncts (1933, p. 195):
E n d o c e n t r i c c o n s t r u c t i o n s a r e of t wo k i n d s , co-ordinate ( or serial) a n d subordinative ( o r
THE CONCEPT OF PHRASE STRUCTURE 341
attributive). In the former type the resultant phrase belongs to the same form-class as two
or more of the constituents.
Bloch (1946: pp. 228-230) analyzes Japanese coordinate noun phrases
in almost the same way, treating "t he head [as] consist[ing] of a series of
two or more nouns, each noun with or without one or more modifiers
preceding, and each noun, or each except the last, followed by a CON-
J UNCTI VE PARTI CLE" . There is also evidence for this sort of analysis
in Harris' s [1951, p. 289) description of Moroccan Arabi c coordination.
While, on the ot her hand, he says that "Any morpheme class or sequence
plus u plus an equivalent morpheme class or sequence equals the mor-
pheme class or sequence itself", 2L he goes on to add that "When two or
more N ~ occur with Va, the Va contains the plural mor pheme. . . - . 22 The
addition of the words or mo r e suggests that Harris intended his statement
of coordination to allow unbounded sequences of conjuncts rather than
merely two. z3
Insofar as IC analysis provided for unbounded branching, then, Chore-
sky' s formalization was off the mark. The use of the Kleene star in GPSG
thus represents a return to structuralist roots but it is nonetheless a crucial
departure from the model that Chomsky was arguing against in the first
place. Strictly speaking, a grammar with rules of the form A - - ~ X A * Y is
not a phrase structure grammar in Chomsky' s sense. In current syntactic
literature no distinction is made between grammars which use regular
expressions and grammars which do not. But while the use of regular
expressions has no effect on weak generative capacity, the fact that it does
affect strong generative capacity means that the question of whether such
devices are available is highly significant in the context of disputes over
the theoretical adequacy of different models.
Di s c o n t i n u o u s Cons t i t uent s
With regard to discontinuous constituents, the story is partly though not
exactly analogous. Discontinuous constituents were also featured in many
IC analyses. A good example is Bloomfield' s analysis of parentheticals
(1933, p. 186): 24
21 The morpheme u is a coordinating conjunction ('and').
;2 The symbol N 4 denotes a noun phrase, and Va stands for verb suffixes marking subject-
verb agreement.
23 It may be worth recalling that Syntactic Structures (Chomsky, 1957) also provided (trans-
formational) rules only for binary coordination.
24 Bloomfield even allowed for discontinuous words, although he regarded these as abnormal
(pp. 180-18t).
342 AL E XI S MANAS T E R - R AME R AND MI C HAE L B. KAC
Pa r e n t h e s i s is a var i et y of par at axi s in whi ch one f or m i nt er r upt s t he ot her ; in Engl i sh t he
par ent het i c f or m is or di nar i l y pr e c e de d and f ol l owed by a pause- pi t ch: I s a w t he b o y [ , ] I
me a n S mi t h ' s b o y [ , ] r u n n i n g ac r os s t he s t r eet [ . ] In a f or m like Wo n ' t y o u p l e a s e c o me ? t he
p l e a s e is in a c l os e par ent hes i s , wi t hout pause- pi t ch.
Likewise, in his work on Japanese syntax, Bloch (p. 229) finds an adverb
like ma t a ' again' inside a discontinuous constituent consisting of a series
of conjoined nouns in certain coordinate constructions.
Even more telling is the fact that the problems handled by Chomsky
(1957, pp. 39-41 and passim) in terms of the affix hopping transformation
in English seem analogous to phenomena in Yokut s and German described
by Harris (1951, pp. 165-66) in terms of discontinuous constituents. The
Yokut s construction indicates uncertainty and is formally characterized by
the cooccurrence of the independent element na' as, and a verb stem with
the suffix -al , e.g., x a t x a t - a l na'as. ~ na'as, x a t x a t - a l . Since na'as, and - al
always occur together, Harris proposes to analyze the two elements as a
discontinuous morpheme. In German, Harris analyzes the ge- prefix of
perfect participles and certain adjectives as forming a part of discontinuous
morphemes together with the - et and - e n suffixes, in forms like g e e i g n e t
' suitable' and g e f a n g e n ' captive' . In fact, Harris considered such an analy-
sis for the English "affix hopping" phenomena, but dismissed it on the
grounds that - en can occur not only with h a v e but also with be and
elsewhere (p. 214). This objection would, of course, still leave open the
possibility that in English the auxiliary stems and the cooccurring suffixes
would form complex discontinuous constituents (rather than simple dis-
continucus morphemes).
Chomsky (1957, p. 41, no. 6) also not ed the possibility of treating the
English facts in terms of discontinuity but rejected it out of hand, and left
the very possibility of discontinuity out of his model of phrase structure.
The subsequent history of discontinuous constituents has been quite differ-
ent from that of unbounded branching, however. They were explicitly
incorporated into the extensions of Chomsky' s PSG introduced by Yngve
(1960) and Harman. Moreover, this was the only point on which Harman
specifically argued that Chomsky' s model of PSG was not faithful to the
IC analysis. 2s However, discontinuous constituents were not incorporated
into TG. They were also left out of GPSG, and have not become part of
the popular conception of phrase structure. Only in recent years have
some syntacticians been showing signs of readiness to bite the bullet on
25 I nt er est i ngl y, Ha r ma n' s aut hor i t y for IC anal ysi s was Post al ' s t hen unpubl i s hed book
(1964a), and Ha r ma n' s r eadi ng was qui t e cor r ect , as we have shown.
THE CONCEPT OF PHRASE STRUCTURE 343
this matter, e.g., Bach (1981), McCawley (1982), Kac (1985a), Ojeda
(1987), and Huck and Ojeda (1987).
C r o s s - c l a s s i f i c a t i o n
The situation with cross-classification is different still. Recent conceptions
of phrase structure, such as GPSG, make extensive and crucial use of
complex symbols, they have been part and parcel of transformational
grammar since 1965, and they had been advocated by Harman in his
defense of phrase structure two years before. To get the history straight,
let us start, for the moment, with Harman, who pointed out that a PSG
using complex symbols would stand up to many of the arguments Chomsky
had advanced against PSG's with only simplex symbols. Chomsky (1965,
1966), while conceding this up to a point, rejoins by noting that, precisely
because the use of complex symbols distinguishes Harman' s model nontri-
vially from Chomsky's own definition of PSG, the advantages of Harman' s
grammars, if any, are irrelevant to the question of the adequacy of PSG.
Something like a replay of this scenario, but with a difference, took place
with Gazdar' s somewhat heated polemic against Chomsky's contention
that PSG's with complex symbols are not PSG's but form a distinct class
of grammars, with some relation to transformational grammars. Gazdar
rejects Chomsky' s argument on the grounds that CFG' s using complex
symbols (such as GPSG's) are weakly and strongly equivalent to simple
CFG' s, whereas TG' s are vastly more powerful in both respects. Since,
as noted, complex symbols were adopted in TG in the aftermath of the
Harman-Chomsky debate, the issue here cannot be the need for this
device; nor is there much disagreement about the additional descriptive
power that it imparts to a formalism. Rather, what seems to be at issue
is the propriety of calling a grammar that uses it a PSG.
However, the issue is not just definitional, because ultimately PSG was
supposed to be a complete formalization of a well-articulated descriptive
model, that of Bloomfieldian IC analysis, and this raises the question
whether complex symbols were used in IC models. This was n o t claimed
by Harman, 26 and Chomsky must have assumed the contrary. That this
point i s crucial can be seen by considering that, if complex symbols had
n o t been employed by the structuralists, then Harman' s would have been
no defense of phrase structure, though it might have been an interesting
26 Harman has maintained throughout that a grammar with complex symbols could be
viewed as an abbreviation of a CFG and that to that extent he had not introduced a crucial
modification of Chomsky's notion of PSG.
344 AL E XI S MANAS T E R - R AME R AND MI C HAE L B. KAC
demonstration of the utility of a new model of grammar, distinct from
bot h TG and PSG. If, on the ot her hand, he was formalizing something
that had been there, even implicitly, in the IC model, then it is Chomsky' s
purport ed refutation of phrase structure that loses much of its force.
A case can be made, however, that the structuralists made use of the
concept of complex symbol, though usually without any special notation.
For example, Bloch in his IC analysis of Japanese repeat edl y makes use
of the implicit convention that a statement, say, about ' predicates' refers
to 'final predicates' and 'non-final predicates' as well as 'plain predicates'
and ' polite predicates' (1946, pp. 206, 213-218). 27
Likewise, Bloomfield in his discussion of English grammatical categories
clearly presupposes that statements about, for example, ' adjectives' cover
' descriptive adjectives' and 'limiting adjectives' (1933, pp. 202-203). Fur-
thermore, Bloomfield was clearly aware of cross-classification in English,
e.g. (1933, p. 203):
Ou r l i mi t i ng adj ect i ves fall i nt o t wo s ub- cl as s es of determiners and numeratives. The s e
t wo cl asses ha ve s ever al s ubdi vi s i ons a nd ar e cr os s ed, mo r e o v e r , by s ever al ot he r l i nes of
cl assi fi cat i on.
Indeed, it appears that Bloomfield was one of the first writers on language
to give explicit recognition to the phenomenon of cross-classification
(1933, p. 269):
For m- cl as s es ar e not mut ua l l y excl usi ve, but cr oss each ot he r a nd over l ap and ar e i ncl uded
one wi t hi n t he ot he r , and so on. Th u s , in Engl i s h, t he nomi na t i ve expr es s i ons (whi ch s er ve
as act or s) i ncl ude bot h s ubs t ant i ves a nd ma r k e d i nfi ni t i ves (to scold the boys would be
foolish). On t he ot he r h a n d , a mo n g t he s ubs t a nt i ve s are s o me p r o n o u n - f o r ms whi ch, by
over di f f er ent i at i on, do not ser ve as act or s: me, us, him, her, them, whom. On e gr oup of
s ubs t ant i ves , t he ge r unds (scolding), bel ongs to a f or m- cl as s wi t h i nfi ni t i ves a nd wi t h ot he r
ver b f or ms , in s er vi ng as he a d for cer t ai n t ypes of modi f i er s, s uch as a goal (scolding the
boys). For t hi s r e a s on a s ys t e m of par t s of s peech in a l anguage like Engl i s h c a nnot be set
up in any ful l y s at i s f act or y wa y . . .
Likewise, in Bloch' s analysis of Japanese, we find perfectly clear discussion
of cross-classification involving final/non-final and polite/plain predicates,
with all four possibilities occurring (pp. 216-217).
The fact that categories called aA and bA can be jointly referenced by
A would appear to be such a basic fact of normal English usage - and IC
analyses were written in English rather than in a symbolic metalanguage
- that it would never have occurred to the structuralist writers that final
27 Pos t al mi s u n d e r s t o o d t hi s and t r i ed t o expl ai n s uch cas es as i ns t ances of cont ext - s ens i t i vi t y,
t he i dea be i ng t hat t he cat egor y predicate is r ewr i t t en as final-predicate wh e n d o mi n a t e d by
final-clause a nd as non-final-predicate wh e n d o mi n a t e d by non-final-clause.
THE CONCEPT OF PHRASE STRUCTURE 345
a n d n o n - f i n a l p r e d i c a t e s , s ay, we r e not b o t h o f t h e m p r e d i c a t e s . Th e us e
o f f o r mu l a i c n o t a t i o n a p p e a r s t o ha ve b e e n i n t r o d u c e d b y Ha r r i s , wh o
e xpl i c i t l y a l l ows c a t e g o r i e s , d e n o t e d b y l e t t e r s , t o s u b s u me ( s u b ) c a t e g o r -
i es, d e n o t e d b y t h e s a me l e t t e r s wi t h v a r i o u s s u b s c r i p t s ( 1946, p. 167):
We may distinguish several sub-classes such as those listed below, while V without any
subclass mark will be used to indicate all the subclasses together.
F u r t h e r mo r e , as i s r a t h e r we l l k n o wn , Ha r r i s al s o i n t r o d u c e d t he s y s t e m
o f a na l ys i s whi c h was r e i n t r o d u c e d b y Ch o ms k y (1970) a n d ha s c o me t o
be k n o wn as X- bar theory, a n d wh o s e wh o l e p o i n t is t ha t s y n t a c t i c c a t e g o r -
i es a r e c o mp l e x . 2s
Li near Order
I n P S G' s , c o n s t i t u e n t s a r e c r e a t e d a n d o r d e r e d by t he s a me r ul e s , so t ha t
a c o n s t i t u e n t d o e s n o t exi s t a p a r t f r o m i t s p o s i t i o n wi t h r e s p e c t t o i t s s i s t e r
c o n s t i t u e n t s . Th e a c t u a l I C a p p r o a c h wa s q u i t e d i f f e r e n t , h o we v e r . Th e
I C' s of a c o n s t r u c t i o n di d n o t h a v e t o be o r d e r e d at al l - i n d e e d , f i xed
wo r d o r d e r wa s t r e a t e d as o n e o f a n u mb e r o f p o t e n t i a l f o r ma l a t t r i b u t e s
of a c o n s t r u c t i o n ( Bl o o mf i e l d , 1933; Ho c k e t t , 1954). Spe c i f i c a l l y, a c c o r d -
i ng t o Bl o o mf i e l d ( 1933, pp. 162- 206) , e v e r y c o n s t r u c t i o n is c h a r a c t e r i z e d
by a s et of taxernes, t h a t i s, f o r ma l a t t r i b u t e s , o f whi c h he d i s t i n g u i s h e s
s e v e r a l k i n d s . On e o f t h e s e i n ( l i n e a r ) o r d e r . 29 Cr u c i a l l y , t a x e me s of o r d e r
a r e n o t a l wa ys r e q u i r e d ( Bl o o mf i e l d , 1933, p. 197):
An example of a taxeme of order is the arrangement by which the actor form precedes the
action form in the normal type of the English actor-action construction: John ran. In
languages which use highly complex taxemes of selection, order is largely nondistinctive and
connotative; in a Latin phrase such as pater amat filium ' the father loves the son', the
syntactic relations are all selective (cross-reference and government) and the words appear
in all possible orders (paterfilium amat, filium pater arnat, and so on), with differences only
of emphasis and liveliness.
As t he La t i n e x a mp l e i l l us t r a t e s , a c o n s t r u c t i o n ( i n t hi s c a s e a c l a us e ) ma y
b e d e f i n e d wi t h o u t a n y r e f e r e n c e t o l i n e a r o r d e r . Sy n t a c t i c a l l y , pat er amat
f i l i um a n d pat er f i l i um amat a r e , f or Bl o o mf i e l d , t he s a me c l a us e . Thi s
a p p r o a c h is a mp l y e x e mp l i f i e d i n di s c us s i ons o f t he I C a na l ys i s of l a n g u -
28 Postal (1964a), writing a few years earlier, could make no sense of Harris's proposals and
entertained the possibility that symbols such as N 3 and N 4 should be treated as simplex and
hence unrelated categories, like the N and NP of (pre-X-bar) transformational grammar!
29 The others are modulation (which has to do with to suprasegmentals), phonetic modific-
ation (referring to sandhi processes), and selection (which subsumes agreement, government,
and categorial information).
346 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
ages with a variety of word orders, including Latin, English, German
and French (Bloomfield, 1933, pp. 197ff.), Japanese (Bloch, 1946), and
Menomini (Bloomfield, 1962, pp. 437ff.). An example from Bloch is the
statement that "It]he relative order of clause attributes is not determined
by their type: an adverbial phrase, as such, can precede or follow a
relational phrase" (1946, p. 218).
The difference between IC analysis and PSG, however, has had a differ-
ent history from the three discussed above. It was not highlighted by Harris
in his morpheme-to-utterance work (but see the discussion of Hidatsa in
Harris, 1946), or in either of the neo-PS proposals (Harman' s or the early
versions of GPSG). To be sure, recent versions of GPSG (e.g., Gazdar
et al., 1985) separate the immediate dominance and linear precedence
components of rules. However, it may be that a better formalization of the
IC approach would be in terms of the proposal, adumbrated by Manaster-
Ramer (1976), that sentences are partially ordered sets, rather than in
ID/LP string terms, since the former, unlike the later, allows scrambling
across constituent boundaries, which seems to have been permitted in the
IC model as well.
Null Elements
Finally, we come to null (zero) elements. The IC grammarians appear to
have used this device more sparingly in syntax than in morphology, but
have nevertheless left clear instances of it. For example, Bloomfield refers
to it in connection with zero anaphora in English sentences like Mary
dances better than Jane (1933, p. 252):
We can describe this latter type by saying that (after as and t han) an actor (Jane) serves as
an anaphoric substitute 3 for an actor-action expression (Jane dances) , or we can say that
(after as and t han) a zero- f eat ure serves as an anaphoric substitute for a finite verb expression
accompanying an actor expression.
Similarly, he spoke of a zero serving as a 'relative substitute' in construc-
tions like the man I saw, the house we lived in, and the hero he was (1933,
p. 263). This should perhaps come as no surprise given Bloomfield's well-
known veneration of the first grammarian known to use zeroes, Pfin.ini. A
similar example comes from Bloch's Japanese syntax, involving coordinate
structures without explicit conjunctions (p. 229):
30 It may be useful to point out that Bloomfield used the term ~substitute' to refer to what
would today be called "pro-forms' .
THE CONCEPT OF PHRASE STRUCTURE 347
Many sentences contain a Series of two or more nouns with no particles between them, but
with each noun, or each except the last, followed by a pause; the last noun is followed by
a referent particle, 31 the copula, or the predicate of the clause.
. . To cover sequences of this type, we posit a fifth conjunctive particle, phonemically zero,
with the same syntactic functions as the four already mentioned. The sentence just cited
might be written H6n O, zassi O, siBbuB o, katta, 32 with 0 (zero particle) syntactically
equivalent to the particle to 'and'.
Choms ky al so l eft t hi s aspect of I C anal ysi s out of his defi ni t i on of
CFG' s and CSG' s ( e. g. , 1956, 1959, 1963). Thi s mi ght a ppe a r t o be
cont r adi ct ed by s ome exampl es of phr as e st r uct ur e rul es gi ven in Choms ky
(1957, pp. 29 (fn. 3), 39, 111), wher e a mo r p h e me wr i t t en ' 0' appear s in
pl aces wher e not hi ng is pr onounc e d ( " t he mo r p h e me whi ch is si ngul ar f or
nouns and pl ur al f or ve r bs " ) . Howe ve r , it seems cl ear t hat , as f ar as t he
synt ax is concer ned, t hese are t r eat ed as act ual mor phe me s . I t is onl y
t he ( t r ans f or mat i onal ) mo r p h o p h o n e mi c c ompone nt whi ch ' ha ppe ns ' t o
real i ze t he m as zer oes. Thi s a ppe a r s t o be di f f er ent f r om Bl och' s t r eat -
ment , whi ch does not s eem t o call f or a s epar at e mo r p h o p h o n e mi c rul e
t o del et e t he zer o, but r a t he r t r eat s t he null r eal i zat i on as a fact of synt ax.
I n any event , Bl oomf i el d and Bl och pos t ul at ed zer o el ement s mor e f r eel y
in t hei r I C anal yses t han Choms ky di d in his phr ase- st r uct ur al bas e c ompo-
nent , i ncl udi ng cases whi ch t r ans f or mat i onal gr a mma r i a ns woul d f or year s
consi der t o r equi r e t he use of synt act i c del et i on t r ans f or mat i ons .
Choms ky di d not al l ow expl i ci t null pr oduct i ons t o be par t of GS G ( i . e. ,
PSG) or CFG, t hough, of cour se, t hey ar e necessar y in t ype- 0 gr ammar s .
On t he ot her hand, t hey wer e per mi t t ed in t he phr as e st r uct ur e model of
Ha r ma n and in GPSG, and t hey have al so been all but uni ver sal l y r ecog-
ni zed in mor e r ecent defi ni t i ons of CFG (fol l owi ng t he Bar - Hi l l el et al.
(1961) defi ni t i on of what t hey cal l ed ' si mpl e phr as e st r uct ur e gr a mma r ' )
Co n s e q u e n c e s f o r t he Phr as e St r uct ur e Co n t r o v e r s y
On all five poi nt s ( unbounde d br anchi ng, di scont i nui t y, cross-classific-
at i on, t he s epar at i on of l i near f r om hi er ar chi cal i nf or mat i on, and null
el ement s ) , we find t hat t he act ual I C syst ems e mpl oye d devi ces t hat wer e
excl uded f r om Choms ky' s defi ni t i on of PSG. Mor e ove r , t he fact t hat
pr ovi si on f or di scont i nuous const i t uent s had be e n ma de in t he I C model
was acknowl edged by Post al (1964a, pp. 67-70 and passi m) and Choms ky
(1957, p. 41, n. 6; 1965, p. 210, n. 4; 1966, p. 39, n. 15), and Post al also
3i What would today probably be labeled a 'case particle".
32 Translation: '[I] bought a book and a magazine and a newspaper'.
348 ALEXIS MANASTER-RAMER AND MICHAEL B. KAC
ment i ons unbounde d br anchi ng ( pp. 23-24 and passi m) . To be sure, Post al
(p. 78) ar gued t hat t hese aspect s of st ruct ural i st synt ax had not been
f or mal i zed by t he st ruct ural i st s t hemsel ves, and t hat t r ans f or mat i onal
g r a mma r pr ovi ded t he onl y known f or mal account of t he p h e n o me n o n in
quest i on:
Thus the claim that Chomsky's notation of PSG reasonably and correctly formalizes the
immediate constituent approach to linguistic analysis as developed in America appears well-
founded. The only aspects of description utilized by these authors which are not formalized
by PSG, namely, discontinuities and perhaps unbounded branching, do receive an apparently
correct formalization in TG and this is indeed the only such formalization known.
Howe ve r , t he wor k of Yngve and Ha r ma n in t he ear l y 1960' s s howed
t hat di scont i nui t y coul d be f or mal i zed per f ect l y well in non- t r ans f or ma-
t i onal t er ms. A si mi l ar t r e a t me nt of unbounde d br anchi ng is ma de possi bl e
by t he use Kl eene' s (1956) r egul ar expr essi ons (in par t i cul ar ones wi t h t he
concat enat i on cl osure, or Kl eene st ar, oper at or ) as r i ght - hand sides of CF
pr oduct i ons. 33 Mor e ove r , since no c ompone nt of st ruct ural i st linguistics
had been f or mal i zed, and since Choms ky was cl ai mi ng t o pr es ent pr eci sel y
such a f or mal i zat i on in t he f or m of PSG, surel y it fol l ows t hat t he shor t -
comi ngs of PSG did not refl ect on t he st r uct ur al i st model s. Thi s be c ome s
even cl ear er when we consi der t hat no ment i on was ma de of t he ot her
ways in whi ch PSG di f f er ed f r om I C t heor y. In fact , t o t he ext ent t hat
st r uct ur al i st model s di d cont ai n devi ces appar ent l y capabl e of handl i ng
many, if not all, of t he p h e n o me n a whi ch t r ans f or mat i ons wer e i nt r oduced
to descr i be, t her e woul d s eem to be s ome quest i on about t he pr i ma faci e
case t hat was ori gi nal l y of f er ed f or t r ans f or mat i onal gr a mma r .
Al l this ma ke s it cl ear t hat t he issue of t he adequacy of t he I C model
and t hat of PSG ar e qui t e i ndependent . Choms ky' s ar gument s may di spose
of PSG as a pl ausi bl e candi dat e f or a t heor y of huma n l anguage synt ax,
but , i nsof ar as t hey i gnor e var i ous c ompone nt s of Bl oomf i el d' s, Bl och' s,
and Har r i s ' s model s, t hey have little to say about t he adequacy of t he
l at t er. Of cour se, t he quest i on of exact l y how good t he st ruct ural i st model s
wer e and what we have yet t o l ear n f r om t hem is ver y much open. I t
coul d be, f or exampl e, t hat Choms ky was ri ght in cl ai mi ng t hat any
a t t e mpt to i ncl ude such devi ces as di scont i nui t y, in a PSG l eads onl y t o
" a d hoc and frui t l ess el abor at i on" (1957, p. 41, n. 6). Yet mos t if not all
33 The use of such productions does not alter the weak generative capacity of CFG's, because
the CFL's are closed under the operations involved, namely, union, concatenation, and (in
the case before us) concatenation, or Kleene, closure. These closure facts had been published
by Bar-Hillel et al. (1961) and Chomsky (1963, p. 380).
THE CONCEPT OF PHRASE STRUCTURE 349
of these elaborations have since found their way into transformational and
post-transformational syntax, whereas transformations - that other great
elaboration of phrase structure - are no longer as widely accepted among
generative grammarians as they were in the 1960's and the early '70's.
Moreover, if such devices were indeed ill-advised, then the case against
structuralist syntax would have been strengthened by including them in
the PSG formalism, and then showing just what was wrong with them,
rather than leaving them out of the account.
At the same time, Chomsky's negative judgment about PSG (that is, the
model he himself defined) has never really been subjected to a substantive
challenge. To see that this is indeed so, consider the fact that every time
that something labeled as a PSG has been advocated as a plausible model
of human syntax, the defenders of phrase structure have enriched the
formalism in ways designed to handle precisely the difficulties that Chom-
sky had identified as insuperable given the simple model of PSG. If
Chomsky had been wrong, there would have no been no need for null
productions, regular expressions, complex symbols, discontinuous con-
stituents, ID/LP, or any number of other devices employed in recent
syntactic theories.
This puts the controversy over phrase structure in a new light. The
debate began with Chomsky's attempt to show the inadequacy of the IC
model. For this issue, the relevant question is not whether Harman or
Gazdar et al. have come up with notational variants of PSG or of TG or
with something different still; it is rather whether Chomsky's proposed
formalization corresponded faithfully to the structuralist practices. If, for
example, we agree that Harman came closer to capturing the actual IC
ideas by allowing for complex symbols, discontinuous constituents, and
deletions than Chomsky did by excluding these, then it was Chomsky and
not Harman who clouded the issue with terminological equivocation. That
is, Chomsky's use of PSG as a stand-in for IC theory, in spite of the major
differences between them, amounted to the introduction of a new and
private notion of IC analysis, divorced from the established usage. In sum,
much of the actual argument that has taken place over the adequacy of
phrase structure has been of little genuine import, for precisely the reasons
given by Chomsky in his response to Harman, though clearly not solely,
or even principally, through any fault of Harman' s.
Interestingly, while Harman (1966) specifically purports to defined PSG
as defined by Chomsky, his discussion leaves some doubt as to whether
he is really ready to accept simple CFG, since he is still committed to
complex symbols, discontinuous constituents, and null productions, but
considers them irrelevant on the grounds that rules with complex symbols
350 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
are abbr evi at i ons of sets of rul es with simple symbol s, and t hat gr ammar s
with di scont i nuous const i t uent or null pr oduct i ons can always be r ewr i t t en
wi t hout such rules. It appear s t hat Ha r ma n is maki ng t he same ki nd of
mi st ake t hat advocat es of GPSG woul d l at er fall i nt o by argui ng t hat ,
since ever y GPSG can be ' compi l ed' i nt o a CFG, t her e is a st rong equi val -
ence bet ween t he t wo classes of grammars. 34 The pr obl em her e is t hat , if
t he t wo model s wer e i ndeed st rongl y equi val ent , t hen it is har d t o see
why t he mor e compl ex GPSG formal i sm was i nt r oduced in t he first pl ace.
The same ar gument applies, mut at i s mut andi s, t o Har man' s posi t i on.
We woul d like t o add t hat an ar gument over whose defi ni t i on of a t er m
is t o be used is not mer el y, as Gazdar (1982) woul d have it, an exerci se
in t er mi nol ogi cal i mperi al i sm; if t he part i es t o a cont r over sy do not agr ee
on what t he t er ms of t he debat e mean, t her e is not hi ng left to debat e.
But , while we agr ee wi t h Choms ky on this poi nt , t he sword, as we have
shown, cuts bot h ways.
3. WHAT P S G IS TODAY: UNI SI NI STRALI TY
We now come t o a deeper quest i on: Is t her e af t er all a not i on of phr ase
st ruct ure, distinct bot h f r om Chomsky' s nar r ow f or mal defi ni t i on and
f r om Bl oomfi el d' s br oader descri pt i ve pract i ce, which has gai ned gener al
cur r ency among cont empor ar y syntacticians, whi ch is bot h coher ent and
nont ri vi al , and whi ch can be defi ned explicitly and t hus made available
f or seri ous discussion? We t hi nk t hat t her e is, t hat is, t hat t he compl i cat ed
hi st ory we have sket ched above has engender ed a popul ar concept i on of
human l anguage synt ax, i dent i fi ed by most wi t h phr ase st r uct ur e gr ammar
but in fact sorel y in need of an unambi guous defi ni t i on as well as an
unambi guous label. We suggest t hat t o count as a PSG in this i nt ui t i ve
sense c ommon t o most syntacticians, it is necessar y and sufficient f or a
gr ammar t o have t he fol l owi ng charact eri st i cs:
(i) The rules of t he gr ammar are all of t he f or m (X, Y), wher e X
is a single gr ammar symbol and Y is a (possibly null) string of
gr ammar symbol s or a r egul ar expressi on, and bot h X and t he
symbol s in Y may be ei t her si mpl ex or compl ex symbols.
(ii) Each such or der ed pai r is i nt er pr et ed in one of t hr ee ways:
as an admissibility condi t i on on t rees allowing a node l abel ed
34 Rather than stating that the equivalence obtains, trivially, between CFG's and the "com-
piled' object grammars generated by GPSG's (see below).
THE CONCEPT OF PHRASE STRUCTURE 351
X t o di r ect l y domi nat e a string of nodes l abel ed (in or der )
ei t her by Y (if Y is a string of gr ammar symbol s) or by some
string in t he r egul ar set denot ed by Y (if Y is a r egul ar ex-
pr essi on) ,
or as a pr oduct i on rul e allowing t he rewri t i ng of t he symbol X
in a sent ent i al f or m 35 by Y (i f Y is a string of gr ammar symbol s)
or by some string in t he r egul ar set denot ed by Y (if Y is a
r egul ar expr essi on) ,
or as a t r ee f or mat i on rul e allowing nodes l abel ed by t he sym-
bols of Y (if Y is a string of gr ammar symbol s) or by t he
symbol s of some string in t he r egul ar set denot ed by Y (if Y is
a r egul ar expressi on) t o be put under a hi t her t o t er mi nal node
X.
Thi s defi ni t i on allows us t o t ake gr ammar s t o oper at e ei t her in t er ms
of string or t r ee rewri t i ng or in t er ms of node admissibility. Thi s cat hol i ci t y
seems t o be r equi r ed gi ven t hat all t hr ee appr oaches are to be f ound in
cur r ent l i t er at ur e.
The onl y significant rest ri ct i on on t hese gr ammar s, t hen, is t hat t her e
be onl y one symbol on t he l ef t - hand side of each rul e, and we t ur n t he
spot l i ght on this f eat ur e by calling such gr ammar s uni s i ni s t r al . To be sure,
t her e has been occasi onal ment i on of cont ext -sensi t i ve rul es within t he
GPSG l i t er at ur e, but it seems cl ear to us t hat t hese have not ent er ed t he
mai ns t r eam of synt act i c t hought associ at ed with t he concept of phr ase
st ruct ure.
It shoul d be not ed t hat we allow t he nont er mi nal vocabul ar y t o be an
infinite (specifically, a recursi vel y enumer abl e) set of symbol s and t hat t he
set of rul es of t he gr ammar may likewise be infinite (recursi vel y enumer -
abl e). Of cour se, an infinite gr ammar cannot be pr es ent ed in t he usual
way. Howe ve r , it woul d appear t hat t he popul ar concept i on of phr ase
st r uct ur e whi ch we are t ryi ng t o expl i cat e allows t he PSG t o be itself
der i ved by a hi gher - or der gr ammar ( met agr ammar ) . The der i ved gr ammar
( obj ect gr ammar ) may t hen be infinite wi t hout vi ol at i ng t he intuitive re-
qui r ement t hat , f or somet hi ng t o be a gr ammar , it must be finitely specifi-
able. The fact t hat such a gener al i zat i on of t he concept of ( phr ase struc-
t ur e) gr ammar is i ndeed necessar y t o account f or t he st at e of t he field is
cl ear f r om t he wi del y- accept ed claims advanced by Gazdar (1982) and
35 A sentential form is a string derived from the initial symbol. Thus, it may still contain
nonterminals and hence not be generated by the grammar.
352 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
ot hers about the weak and strong equivalence of their GPSG' s to CFG' s.
Since a GPSG met agr ammar derives an object gr ammar rat her t han a
l anguage, these claims can onl y be i nt er pr et ed as referri ng to the sets of
strings and trees gener at ed by these object grammars. Mor eover , while
t here is general agr eement t hat , for linguistic reasons, we shoul d const rai n
such model s so as to guar ant ee the finiteness of object grammars, this
very fact shows t hat an infinite object gr ammar is concept ual l y consistent
with the basic model (and apparent l y also wi t h t he early version of GPSG
of Gazdar and Sag, 1980).
It mi ght seem t hat allowing infinite grammars trivializes the not i on of
unisinistral gr ammar , but we hast en to not e (a) t hat we are not advocat i ng
unisinistral gr ammar ei t her as a cont ri but i on to t he st udy of formal systems
or as a t heor y of human l anguage but rat her as an at t empt at the faithful
represent at i on of an implicit not i on of PSG t hat is al ready wi despread in
moder n syntactic t heory, and (b) t hat even such an apparent l y unrest ri ct ed
model of gr ammar may be quite limited in human l anguage applications.
An exampl e to which we shall ret urn is di scont i nuous const i t uent s, which
unisinistral grammars clearly do not allow, i nasmuch as Y is ei t her a
(cont i nuous) string of symbols or variables, or a regul ar expression denot -
ing a set of such strings.
Ther e are t hree ways in which our unisinistral grammars differ from
cont ext -free grammars as defi ned by Chomsky. First, we admi t regul ar
expressions, t hus allowing unbounded branchi ng. Second, we permi t com-
plex symbols. Thi rd, we allow t he (object) gr ammar to be infinite. All
t hree of these factors seem to have become an integral part of the current
concept i on of phrase st ruct ure, and leaving t hem out woul d have ill served
our goal of capt uri ng t hat concept i on in explicit terms. It shoul d be not ed
t hat the first two of these factors, unbounded branchi ng and compl ex
symbols, appear to have been present in the original Bl oomfi el di an con-
cept i on of IC analysis, as we have argued above. We have not i ncl uded
any of the ot her facets of Bl oomfi el d' s model t hat were left out of Chom-
sky' s phrase st ruct ure model , because t hey do not appear to have become
assimilated to the popul ar concept i on of phrase structure which we are
seeking to describe.
We have t hus del i berat el y chosen to exclude some formal devices which
have been gaining currency, such as the separat i on of rules for i mmedi at e
domi nance and l i near precedence, explicit i ncorporat i on of di scont i nui t y,
and t he like. To the ext ent t hat t hese devices are desirable in t hei r own
right, t hat very fact may poi nt to the linguistic i nadequacy of the popul ar
concept i on of phrase st ruct ure, and allowing for t hem in our decri pt i on
woul d no mor e vindicate the unisinistral model t han the devel opment of
T HE C ONC E P T OF P HR AS E S T R UC T UR E 353
generalized phrase structure grammar did the simple context-free model.
It is important that two kinds of theoretical progress be distinguished: that
which comes from exploiting an existing model in novel ways, and that
which comes from extending or modifying it. Thus, to claim that a model
incorporating, say, discontinuous constituents or the ID/LP rule format,
is 'really' a variant of unisinistral grammar would do nothing more than
confuse the very issues which we have sought to clarify.
4 . T HE C ONS E QUE NC E S AND L I MI T AT I ONS OF UNI S 1 NI S T R AL I T Y
We shall now take up the question of the generative capacity of unisinistral
grammars, beginning with some prefatory remarks on the notion of gen-
erative capacity and its relation to theoretical models of human language.
Almost all the known mathematical results about generative capacity
have to do with weak generative capacity, that is, the set of terminal
strings derived by a grammar. Strong generative capacity, the set of deri-
vations (or else, on other definitions, the set of trees) defined by the
grammar, has been little explored by comparison, and the implications of
results about it often invite misinterpretation. In this connection, two
points stand out as especially significant.
The first is that the Chomsky hierarchy can only be used to measure
weak generative capacity. Although, the Church-Turing thesis implies
that there are no computational devices with greater weak generative
capacity than that of type-0 grammars, the strong generative capacity (i.e.,
the sets of derivations) of type-0 grammars is anything but unrestricted.
For example, no type-0 grammar allows an unbounded-length string of
symbols to replace a given nonterminal in one step in a derivation, whereas
a CFG augmented with productions having regular expressions on the
right-hand side can yield such derivations. Yet such augmented CFG' s
generate only context-free languages, a proper subset of the type-0, or
recursively enumerable, languages. Similarly, no type-0 grammar will ever
define a derivation in which a string of nonterminals like A B is replaced
by a string like a B c in one step, by means of a rewrite rule that rewrites
A as ac. Yet, there exist formal systems which do precisely this but whose
weak generative capacity is equal to that of the CFG' s (for example, the
PSG's of Yngve, 1960, and Harman, 1963).
It should be noted that, since type-0 derivations do not always define
parse trees of the usual sort, we have stated these observations directly
in terms of derivations, i.e. sequences of strings. However, we can also
choose to consider only those type-0 derivations in which a s i n g l e nonter-
354 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
mi nal is r ewr i t t en in a single step. 36 It t hen becomes possible t o associ at e
parse t r ees unambi guousl y with deri vat i ons ( t hough not vice versa), and
t hen we can simply say t hat t r ee sets with unbounded number s of sister
nodes and t r ee 37 sets with di scont i nuous const i t uent s are out si de t he st rong
gener at i ve capaci t y of t ype-0 gr ammar s. Thi s claim does not cont r adi ct
t he obvi ous fact t hat t he deri vat i ons i nduced by a CFG with r egul ar
expressi ons f or m an r. e. l anguage ( pr ovi ded t he deri vat i ons are sui t abl y
coded) . Unde r such an encodi ng, each CF der i vat i on woul d be r epr es ent ed
as a string, and t he set of t hese strings woul d be gener at ed by some t ype-
0 gr ammar , but t hi s t ype-0 gr ammar woul d st i l l be unabl e t o deri ve an
unbounded number of symbol s f r om one nont er mi nal in one step. Mut at i s
mut andi s, t he same is t r ue f or t he case of di scont i nuous const i t uent s.
The second poi nt about st rong gener at i ve capaci t y is t hat it is undef i ned
f or al most all cur r ent model s of human l anguage. The not i on was de-
vel oped f or CSG' s and CFG' s , gr ammar s whose deri vat i ons can be r epr e-
sent ed in t r ee f or m ( Chomsky, 1959 [1965, pp. 128-29]). But most cont em-
por ar y synt act i c t heor i es do not i ncor por at e t he same not i on of der i vat i on
or assign simple t r ee st ruct ures t o t hei r strings. Thi s fact is well illustated
by t he cl ai m of Gazdar (1982, p. 134) t hat t he use of compl ex symbol s in
a CFG does not al t er st rong gener at i ve capaci t y. As was poi nt ed out by
Chomsky (1965, pp. 84ff. ), if each node of a phr ase mar ker is l abel ed by
a compl ex symbol , t hen t he resul t i ng obj ect is no l onger a t ree. The r eason
is t hat t he compl ex symbol s i nt r oduce an ' ext r a di mensi on' owi ng t o t he
fact t hat , while a t r ee is def i ned by rel at i ons of domi nance and pr ecedence
among l abel ed nodes, t he i nt ernal st r uct ur e of t he compl ex symbol s does
not fit this descri pt i on: t he individual f eat ur es maki ng up compl ex symbol s
do not l abel separ at e nodes. Chomsky seems to visualize a phr ase mar ker
r epr esent i ng this ki nd of a st r uct ur e as havi ng an ext r a di mensi on f or t he
component s of t he compl ex symbols.
To be sure, Choms ky not es t hat we coul d r ei nt er pr et compl ex symbol s
as or di nar y node labels (in which case t he phr ase mar ker w o u l d be a t r ee) ,
but t hat this woul d amount to t aki ng each compl ex symbol as a simple
symbol with a l ong name. What Choms ky per haps di d not say qui t e as
cl earl y as he mi ght have, is t hat , if this l at t er pr oposal is adopt ed, t hen
t he resul t i ng phr ase mar ker does not in fact refl ect t he way t he gr ammar
funct i ons. Thi s woul d mean t hat a set of such phrase mar ker t rees coul d
36 If we constrain type-0 grammars to allow only such derivations, weak generative capacity
is not affected (R6vdsz, 1976).
37 On a suitably relaxed definition of a tree that allows the representation of discontinuous
constituents (as in McCawley, 1982).
T HE C ONC E P T OF P HR AS E S T R UC T UR E 355
not be consi der ed t o r epr es ent t he st rong gener at i ve capaci t y of a gr ammar
t hat makes use of t he i nt er nal st r uct ur e of t he compl ex symbols. 3s
It may, of cour se, be possi bl e to defi ne a new and mor e gener al concept
of der i vat i on, whi ch woul d make sense of model s, such as TG or GPSG,
whi ch defi ne l anguages in ot her ways t han convent i onal PSG' s. For exam-
pl e, it may be possi bl e t o r epr es ent t he processes wher eby der i ved rul es
are gener at ed by t he met agr ammar in GPSG, t he processes wher eby
der i ved lexical i t ems are pr oduced in LFG, and so on, as par t of such a
gener al i zed not i on of der i vat i on. Si nce t he i dea of st rong gener at i ve capac-
i t y appear s t o have become i next r i cabl y t i ed up with ar bor eal r epr esent a-
t i ons of t he sort commonl y empl oyed f or CFG' s , we shall r ef er t o any
such mor e gener al not i on as deri vat i onal capacity, and f or t he sake of
s ymmet r y we shall also speak of st ri ng capacity r at her t han weak gener at i ve
capaci t y. We shall likewise r ef er t o der i vat i onal and string equi val ence of
formal i sms.
It shoul d now be cl ear t hat , gi ven a class of gr ammar s, ar gument s f or
or against its use as a model of anyt hi ng, e. g. , human l anguage, shoul d
not conf ound string capaci t y and der i vat i onal capaci t y, since consequences
wi t h r egar d t o one Of t hese may not appl y t o t he ot her . 39 For exampl e, if
we are deal i ng wi t h a non- CF l anguage, we know not onl y t hat we cannot
use a CFG t o descri be it but also t hat we cannot empl oy any gr ammar
t hat is st r i ng- equi val ent to a CFG. Fur t her , since t her e is a string capaci t y
hi er ar chy, namel y, t he Choms ky hi er ar chy, we can say somet hi ng about
wher e we mi ght l ook f or a gr ammar t hat will gener at e t he l anguage. But
if we r equi r e non- CF der i vat i onal capaci t y (e. g. , a t r ee set t hat cannot be
def i ned by a CFG) , t he si t uat i on is qui t e di fferent . Fi rst of all, t he class
of devi ces t hat are der i vat i onal l y equi val ent t o t he CFG' s is pr oper l y
cont ai ned wi t hi n t hat of t he weakl y CF formal i sms. Second, since t her e
is no known der i vat i onal capaci t y hi er ar chy, t her e is little t hat we can say
about wher e t o l ook f or an adequat e gr ammar . If, t o t op things off, we
need a model t hat handl es st ruct ural r epr esent at i ons mor e compl ex t han
3s Choms ky al so not es t hat , whi l e a si mpl e PSG def i nes onl y st ri ct l y Mar kovi an der i vat i ons
(in t he sense t hat t he appl i cabi l i t y of a r ul e de pe nds sol el y on t he most r e c e nt line in t he
der i vat i on) , t he use of compl ex symbol s al l ows access t o ear l i er l i nes in a der i vat i on, j ust as
t r ans f or mat i ons do. Ga z da r (1982) si dest eps t hi s ar gument by poi nt i ng t o t he maj or di ffer-
ences be t we e n TG and CFG wi t h compl ex symbol s. Howe ve r , t he fact t hat t wo model s are
di f f er ent in cer t ai n ways does not pr ecl ude t hei r bei ng i dent i cal , or si mi l ar, in ot her s , and
Choms ky' s obs er vat i on cl earl y shows t hat CFG wi t h compl ex symbol s, like TG, can e nc ode
gl obal , as oppos e d t o l ocal , const r ai nt s.
39 A f or t i or i , we mus t keep bot h of t hes e qui t e di st i nct f r om ar gument s t hat appeal t o
el egance, nat ur al ness, and ot he r i nf or mal not i ons.
356 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
t r ees - whi ch woul d be t he case f or all t he model s t hat have wi de cur r ency
in c ont e mpor a r y synt ax - t hen we find our sel ves ver y f ar i ndeed f r om t he
hel pi ng hand of f or mal l anguage t heor y.
Anot he r i mpor t ant mor al follows: I f we wer e t o t ake CFG' s as our
poi nt of depar t ur e and come up wi t h an ar gument t hat huma n l anguages
r equi r e ri cher st r uct ur al descr i pt i ons t han t hose i nduced by a CFG, but
had no r eas on to de ma nd gr eat er st ri ng capaci t y t han t hat of t he CFG' s ,
t hen we woul d not be j ust i fi ed in adopt i ng a model of g r a mma r t han in
fact is much mor e power f ul in t er ms of st ri ng capaci t y t han t he CFG' s .
By t he s ame t oken, if our poi nt of depar t ur e wer e a model well beyond
t he CFG' s in bot h st ri ng and der i vat i onal capaci t y, and we di scover ed t hat
all huma n l anguages ar e CF as f ar as t he st ri ng capaci t y is concer ned, but
f ound no way t o capt ur e t he st r uct ur al pr oper t i es of huma n l anguages in
si mpl e CF t er ms, t hen we woul d not want to adopt CFG as our model of
gr a mma r . Thes e obs er vat i ons mi ght appar to be so obvi ous as t o be
r edundant , wer e it not f or t he fact t hat , as our discussion has shown, t he
hi st or y of t he ori gi ns of TG shows signs of di sr egar d of t he first poi nt ,
whi l e t he hi st or y of t he at t empt s t o r evi ve PSG r eveal s equal negl ect of
t he second one.
The real i t y of huma n l anguage synt ax, in fact , s eems qui t e close ( t hough
not i dent i cal ) to t he second of t he hypot het i cal si t uat i ons we have de-
scri bed. On t he one hand, we can r ei t er at e our ear l i er obs er vat i on t hat it
is possi bl e f or a f or mal i s m to be as power f ul as desi r ed in t er ms of st ri ng
capaci t y and yet t o be sufficiently r est r i ct ed in der i vat i onal capaci t y (as in
t he case of t ype- 0 gr ammar s ) as t o be usel ess f or any human l anguage
appl i cat i on. 4 At t he s ame t i me, t her e does not s eem t o be any compel l i ng
r eas on t o t ake t he st ri ng capaci t y of human l anguage gr a mma r s much
beyond t hat of CFG' s , even if t he few const r uct i ons whi ch i nduce non-
cont ext f r eeness are as cent r al as ar gued by Rounds et al. (1987) and
Ma na s t e r - Ra me r (1987). Nor ar e t hese concl usi ons cont r adi ct or y, gi ven
t hat st ri ng capaci t y, der i vat i onal capaci t y, and ot her r el evant pr oper t i es
of gr a mma r s ar e l argel y i ndependent of each ot her . The di scover y t hat
t he CFG' s ar e i nadequat e f or huma n l anguage ( Ga z da r and Pul l um, 1985,
and t he r ef er ences ci t ed t her ei n) cannot be t aken to i ndi cat e a need f or
a class of gr a mma r s di fferi ng f r om t he CFG' s mor e t han slightly in t er ms
of st ri ng capaci t y (such as TG) . Si mi l arl y, t he di scover y t hat human l an-
guages a ppr oxi ma t e cont ext - f r ee st ri ng sets cl osel y, if not exact l y (i bi d),
4o For much the same reasons the type-0 grammars or Turing machines are never used for
specifying computer programs.
THE CONCEPT OF PHRASE STRUCTURE 357
does not i nval i dat e the first poi nt , and says virtually not hi ng about the
form of grammars requi red for human languages.
Agai nst the backgr ound provi ded by t hese observat i ons, we now not e
t hat t hough unisinistrality is i ncompat i bl e with overt context-sensitivity, it
does not hi ng to pr event non-cont ext -freeness of the gener at ed l anguage
f r om being achi eved covertly. Unisinistrality does not suffice to limit weak
generat i ve capacity, and this reflects the way t hat the concept i on of phrase
st ruct ure has evolved within moder n syntactic t heory. For exampl e, the
GPSG' s of t he sort descri bed by Gazdar and Sag (1980) appar ent l y genera-
t ed the type-0 (recursively enumer abl e) l anguages, since the const rai nt s
on met arul es proposed at the time allowed the number of rules in the
obj ect gr ammar to be infinite. 4~ To be sure, t here are many such grammars
which no one would want to seriously ent ert ai n as descriptions of any
human l anguage, and the consi derabl e i nt erest of t hat version of GPSG
had to do with the propert i es of onl y a small subset of the range of
allowable grammars. (In this respecL the hi st ory of GPSG is not unlike
t hat of early t r ansf or mat i onal gr ammar . ) Thus, even t hough argument s
about weak generat i ve capacity have on occasion been seen as beari ng on
the quest i on of the adequacy of phrase st ruct ure model s, in real i t y t here
is little if any connect i on.
Perhaps more i nt erest i ng in the cont ext of the debat es about phrase
st ruct ure vs. t r ansf or mat i onal gr ammar is the fact t hat unisinistrality is
not onl y not sufficient to achieve the l i mi t at i on of weak generat i ve capacity
to CFL' s, but not necessary either. It shoul d be not ed t hat some of the
unease about t ransformat i onal gr ammar s and the resulting malaise in
syntactic t heor y in the 1970's and 1980's has had to do with Pet ers and
Ri t chi e' s (1971) pr oof t hat TG' s are string equi val ent to type-0 grammars.
The nont r ansf or mat i onal model s t hat have prol i ferat ed since have fre-
quent l y been her al ded, among ot her things, as solutions to t he pr obl em
of excessive string capacity. However , a t r ansf or mat i onal gr ammar can be
const rai ned in such a way as to generat e less t han the full set of type-0
languages. For exampl e, a TG with a CF base component and an arbi t rary
set of opt i onal t r ansf or mat i ons 42 is guar ant eed to generat e a CF l anguage
if it is strongly st ruct ure preserving, i. e. , if every deri ved st ruct ure is
41 Uszkoreit and Peters (I985) prove this, as well as the fact that, even if we allow only one
variable ranging over an infinite set of strings, in the metarules, we can still get all type-0
languages.
42 If we allow obligatory transformations, the construction outlined here for a CFL-genera-
ting TG will not work, but obligatory transforma}ions have not been used for a number of
years.
358 AL E XI S MANAS T E R - R AME R AND MI C HAE L B . KAC
r equi r ed t o be i dent i cal to some base- gener at ed st ruct ure. The st rong
st ruct ure pr eser vat i on const r ai nt can be enf or ced by an al gor i t hm which
verifies t hat t he out put of each t r ansf or mat i on is equal to some base
st r uct ur e and accept s it onl y in t hat case. Since CFL' s are recogni zabl e,
such a filtering al gor i t hm cl earl y exists, and in fact it is easy to i mpl ement .
To be sure, as Tom Wasow (p. c. ) r emi nds us, it was t aken f or gr ant ed
f or a l ong t i me t hat such a rest ri ct i on on TG' s woul d be empi ri cal l y
i nadequat e ( pr esumabl y because it was bel i eved t hat human l anguages
wer e non- cont ext - f r ee) . 4s We are not arguing, however , t hat such a trans-
f or mat i onal gr ammar woul d be a satisfying model of human l anguage.
Rat her we cont end t hat , even if human l anguages were cont ext - f r ee, this
of itself woul d not be evi dence against t r ansf or mat i onal gr ammar , any
mor e t han a demonst r at i on t hat human l anguages are not cont ext - f r ee
shoul d count as an ar gument against phr ase st ruct ure.
We concl ude this sect i on by ment i oni ng some synt act i c phe nome na
whi ch raise subst ant i ve issues regardi ng unisinistral gr ammar s t hat need
to be addressed. Such phe nome na as f r ee wor d or der , di scont i nui t y, agree-
ment , funct i onal rel at i ons, and r edupl i cat i on appear to resist sat i sfact ory
t r eat ment in unisinistral t er ms even if t he r el evant sets of strings can be
gener at ed. To be preci se, unisinistral gr ammar s, like t ype-0 gr ammar s,
gener at e all t he string sets t hat may be equat ed with t he possible human
l anguages, but this does not mean t hat unisinistral gr ammar s are any mor e
pl ausi bl e as model s of human l anguages t han are t ype-0 gr ammar s.
Al l of t he phenomena ment i oned are ones which it may be i mpossi bl e
t o handl e in a pri nci pl ed and nat ur al way within t he unisinistral f r ame-
work. Just like t ype-0 gr ammar s, t hen, unisinistral gr ammar s may fall far
short of t he goal of describing t he synt ax of human l anguages even if t hey
are per mi t t ed full l at i t ude in t er ms of string capaci t y. If t he t r eat ment of
t hese phe nome na as at best special cases is i nevi t abl e under t he unisinistral
assumpt i ons and yet t heor et i cal l y unjustified, t hen a ver y di f f er ent concep-
t i on of gr ammar will be r equi r ed. Cur r ent t r ends in t he field of synt ax in
fact argue t hat this is t he case. Among t hese we must not e t he I D/ LP
f or mat which di vorces rul es governi ng l i near or der f r om t hose cont rol l i ng
domi nance, t he f or mal i zat i on of di scont i nuous const i t uent s by wrappi ng
oper at i ons, and t he explicit t r eat ment of rel at i onal (funct i onal ) cat egori es.
Of course, as not ed, t wo of t he original IC devi ces t hat had been left out
of Chomsky' s f or mal i zat i on of phr ase st r uct ur e ( unbounded br anchi ng
43 I n s t e a d , c o n s i d e r a b l e e n e r g y wa s d e v o t e d t o r e s t r i c t i n g t r a n s f o r ma t i o n s s o as t o e x c l u d e
n o n - r e c u r s i v e l a n g u a g e s ( s ee Wa s o w, 1978, a n d t h e l i t e r a t u r e c i t e d t h e r e i n ) .
THE CONCEPT OF PHRASE STRUCTURE 359
and cross-classification) have won vi rt ual l y uni versal accept ance. We have,
t her ef or e, i ncl uded t hem in our concept i on of unisinistral gr ammar , since
t hey appear no l onger t o be at issue. But , t o keep t rack of t he issues, we
have kept out of unisinistral gr ammar ot her devi ces whi ch had been par t
and parcel of Bl oomf i el d' s t heor y of human synt ax but which have not
been gener al l y i ncl uded in cont empor ar y concept i ons of phr ase st ruct ure.
4. CONCLUSI ONS
In this paper , we have ar gued t wo mai n points:
(i) Much of t he debat e over t he adequacy of phr ase st r uct ur e
anal yses of human l anguage synt ax has been r ender ed poi nt l ess
by i nconsi st ency in t he way t he t er m p h r a s e s t r uct ur e g r a mma r
has been used. By t he same t oken, however , t he ori gi nal trans-
f or mat i onal i st cri t i que of I C analysis is r ender ed suspect by
t he di screpanci es bet ween t he act ual Bl oomf i el di an model and
Choms ky' s f or mal i zat i on of it in t er ms of PSG.
(ii) Ther e is nonet hel ess a coher ent and subst ant i ve issue which
needs t o be addr essed in cont empor ar y synt act i c t heor y, name-
ly, t he adequacy of a concept i on of phr ase st r uct ur e whi ch we
have t ri ed to expl i ci t at e in t er ms of our not i on of unisinistrality.
In t he cour se of discussing t hese poi nt s we have also sought t o clarify such
cent ral t heor et i cal concept s as gener at i ve capaci t y, compl ex symbol , and
t r ee. We have sought , with our obser vat i ons r egar di ng t he gener at i ve
capaci t y and t he adequacy of t ype-0 and unisinistral gr ammar s, t o exemp-
lify t he i mpor t ant pri nci pl e t hat even an appar ent l y highly st ri ngent restric-
t i on on t he f or mat of gr ammar rules may have no effect on t he string
capaci t y of t he model . Yet at t he same t i me, a f or mal i sm may be unrest ri c-
t ed in t er ms of string capaci t y, but still have no use as a t ool f or anal yzi ng
human l anguages (as in t he case of t ype-0 gr ammar s) . If not hi ng else, this
argues f or caut i on in maki ng and eval uat i ng claims about t he power ,
equi val ence, and adequacy of di f f er ent model s of gr ammar .
REFERENCES
Bach, E.: 1981, ~Discontinuous Constituents in Generalized Categorial Grammar', Proceed-
ings of the Eleventh Conference of the New England Linguistic Socieo,, pp. I-i2.
Bar-Hillel, Y., M. Perles, and E. Shamir: 1961, "On Formal Properties of Simple Phrase
Structure Grammars', Zeitschrift fiir Phonetik, Sprachwissenschaft und Kommunika-
tionsforschung 14, 143-172.
360 ALEXI S MANAS T E R - R AME R AND MI CHAE L B. KAC
Berwick, R. and A. Weinberg: 1982, ' Parsing Efficiency, Comput at i onal Complexity, and
t he Eval uat i on of Grammat i cal Theori es' , Linguistic Inquiry 13, 165-91.
Bloch, B.: 1946, "Studies in Colloquial Japanese II: Syntax' , Language 12, 200-48.
Bloomfield, L.: 1914, An Introduction to the Study of Language, Henry Hol t and Co. , New
York.
Bloomfield, L.: 1933, Language, Holt, Ri nehart , and Wi nst on, New York.
Bloomfield, L.: 1962, The Menomini Language, Yale University Press, New Haven.
Bresnan, J., R. M. Kapl an, S. Pet ers, and A. Zaenen: 1982, ' Cross-serial Dependenci es in
Dut ch' , Linguistic Inquiry 13, 613-35.
Chomsky, N.: 1956, ' Three Model s for the Description of Language' , IRE Transactions on
Information Theory, IT-2, 113-24. Repri nt ed 1965, in R. D. Luce, R. R. Bush, and E.
Gal ant er (eds.), Readings in Mathematical Psychology, Vol. 2, pp. 105-24, John Wiley,
New York.
Chomsky, N.: 1957, Syntactic Structures, Mout on, The Hague.
Chomsky, N.: 1959, ~On Cert ai n Formal Propert i es of Gr ammar s' , Information and Control
1, 91-112. Repri nt ed 1965, in R. D. Luce, R. R. Bush, and E. Gal ant er (eds,), Readings
in Mathematical Psychology, Vol. 2, pp. 124-55, John Wiley, New York.
Chomsky, N.: 1961, ' On the Not i on "Rul e of Gr ammar " ' , in R. Jakobson (ed. ), Proceedings
of Symposia in Applied Mathematics, vol. XII: Sn'ucture of Language and its Mathematical
Aspects, pp. 6-24, Ameri can Mat hemat i cal Society, Providence.
Chomsky, N.: 1962a, "The Logical Basis of Linguistic Theory' , in Preprints of Papers for the
Ninth International Congress of Linguists, August 27-31, 1962, Cambri dge, Mass., pp.
509-74, [Massachusetts Inst i t ut e of Technology?], Cambri dge, MA.
Chomsky, N.: 1962b, ' A Transformat i onal Appr oach to Syntax' , in [A. A. Hill, (ed.)] Third
Texas Conference on Problems of Linguistic Analysis in English, May 9-12, 1958, pp.
124-158, Uni versi t y of Texas Press, Austin.
Chomsky, N.: 1963, ' Formal Propert i es of Gr ammar s' , in R. D. Luce, R. R. Bush, and E.
Gal ant er (eds.), Handbook of Mathematical Psychology, Vol. 2, pp. 323-418, John Wiley,
New York.
Chomsky, N.: 1964, ' The Logical Basis of Linguistic Theory' , in H. G. Lunt (ed. ), Proceed-
ings of the Ninth International Congress of Linguists, Cambridge, Mass., August 27-31,
1962, pp. 914-978, Mout on, The Hague.
Chomsky, N.: 1965, Aspects of the Theory of Syntax, MI T Press, Cambri dge, MA.
Chomsky, N.: 1966, Topics in the Theory of Generative Grammar, Mout on, The Hague.
Chomsky, N.: 1970, ' Remar ks on Nomi nal i zat i on' , in R. Jacobs and P. Rosenbaum (eds.),
Readings in English Transformational Grammar, pp. 184-221, Gi nn, Wal t ham, MA.
Chomsky, N. and G. A. Miller: 1963, ' Int roduct i on to t he Formal Analysis of Nat ural
Languages, in R. D. Luce, R. R. Bush, and E. Gal ant er (eds.), Handbook of Mathematical
Psychology, Vol. 2, pp. 269-322, John Wiley, New York.
Gazdar, G. : 1981, ' Unbounded Dependenci es and Coordi nat e St ruct ure' , Linguistic Inquiry
12, 155-84.
Gazdar, G. : 1982, ' Phrase Structure Gr ammar ' , in P. Jacobson and G. K. Pullum (eds.),
The Nature of Syntactic Represesentation, Reidel, Dordrecht .
Gazdar, G. : 1985, ~Applicability o[ Indexed Grammars to Nat ural Languages' , CSLI Repor t
85-34, Stanford University, Stanford, CA.
Gazdar, G. , and G. K. Pullum: 1982, GPSG: A Theoretical Synopsis, Indi ana University
Linguistics Club, Bl oomi ngt on, IN.
Gazdar, G. and G. K. Pullum: 1985, ~Computationally Rel evant Propert i es of Nat ural
Languages and Thei r Gr ammar s' , New Generation Computing 3, 273-306.
Gazdar, G. , and I. A. Sag: 1980, "Passives and Reflexives in Phrase Structure Gr ammar ' ,
in J. A. G. Groenendi j k, T. Janssen, and M. St okhof (eds.), Formal Methods in the
Study of Language, Mathematical Center Tracts 135, pp. 131-52, Mat hemat i sch Cent rum,
Uni versi t ei t van Amst er dam, Amst erdam.
THE CONCEP T OF P HRAS E S T RUCT URE 361
Gazdar, G. , E. Klein, G. K. Pullum, and I. A. Sag: 1985, Generalized Phrase Structure
Grammar, Harvard Uni versi t y Press, Cambri dge, MA.
Har man, G. H. : 1963, ' Gener at i ve Gr ammar s wi t hout Transformat i on Rules: A Defense of
Phrase St ruct ure' , Language 39, 597-616.
Har man, G. H.: 1966, ~The Adequacy of Cont ext -Free Phrase-St ruct ure Gr ammar s' , Word
22, 276-293.
Harris, Z. S.: 1945, ' Di scont i nuous Mor phemes' , Languages 21, 121-27.
Harris, Z. S.: 1946, ~From Mor pheme to Ut t er ance' , Language 22, 161-83.
Harris, Z. S.: 1951, Methods in Structural Linguistics, University of Chicago Press, Chicago.
Harris, Z. S.: 1962, String Analysis, Mout on, The Hague.
Harri son, M. A. : 1978, Introduction to Formal Language Theory, Addison-Wesley, Reading,
MA.
Hocket t , C. F.: 1954, "Two Models of Grammat i cal Descri pt i on' , Word 10, 210-31.
Hocket t , C. F.: 1958. A Course in Modern Linguistics, Macmillan, New York.
Hocket t , C. F.: 1961, ~Grammar for the Hear er ' , in R. Jakobson (ed. ), Proceedings of
Symposia in Applied Mathematics, vol. XII: Structure of Language and its Mathematical
Aspects, pp. 220-236, Ameri can Mat hemat i cal Society, Providence.
Hopcroft , J. E. and J. D. Ul l man: 1979, Introduction to Automata Theory, Languages, and
Computation, Addison-Wesley, Readi ng, MA.
Huck, G. J. and A. E. Oj eda (eds.): 1987, Discontinuous Constituency [=Syntax and
Semantics, 20], Academi c Press, Orl ando, FL.
Hudson, R. A.: i976, Arguments for a Nontransformational Grammar, University of Chicago
Press, Chicago.
Hudson, R. A.: 1984, Word Grammar, Basil Blackwell, Oxford.
Joshi, A. K~: 1983, ~How Much Context-Sensitivity is Requi red to Provide Reasonabl e
Structural Descriptions: Tree-Adj oi ni ng Gr ammar s' , in D, Dowry, L. Kart t unen, and A.
Zwicky (eds.), Natural Language PatMng: Psycholinguistic, Computational, attd Theoretical
Perspectives, pp. 190-204, Cambri dge University Press.
Kac, M. B.: 1985a, Grammars and Grammaticalio,, unpubl i shed ms., University of
Mi nnesot a.
Kac, M. B.: 1985b, "Constraints on Predicate Coordi nat i on' , Indi ana University of Linguistics
Club, Bl oomi ngt on.
Kleene, S. C.: 1956, "Representation of Event s in Nerve Nets and Finite Aut omat a' , Auto-
mata Studies, pp. 3-42, Pri ncet on University Press, Princeton.
Lamb, S. M.: 1962, Outline of Stratificational Grammar, Uni versi t y of California, Berkeley.
Manast er - Ramer , A.: 1978, ' The Position of t he Ver b in Dut ch and Ger man' , Papers from
the Fourteenth Regional Meeting, pp. 254-63, Chicago Linguistic Society, Chicago.
Manast er - Ramer , A.: 1987, "Dutch as a Formal Language' , Linguistics and Philosophy.
McCawley, J. D. : 1968, ~Concerning t he Base Component of a Transformat i onal Gr ammar ' ,
Foundations of Language 4, 243-269.
McCawley, J. D.: 1982, "Parentheticals and Di scont i nuous Const i t uent Structure, Linguistic
Inquiry 13, 91-106.
Oj eda, A. : 1987, ~Discontinuity and Phrase St ruct ure' , in A. Manast er - Ramer (ed. ), Mathe-
matics of Language, John Benj ami ns, Amst er dam.
Percival, W. K.: 1976, "On the Historical Source of Immedi at e Const i t uent Analysis' , in J.
D. McCawley (ed. ), Syntax and Semantics 7: Notes from the Linguistic Underground, pp.
229-242, Academi c Press, New York.
Peters, S. and R. Ritchie: 1969, "Context-sensitive Immedi at e Const i t uent Analysis - Con-
text-free Languages Revi si t ed' , Proceedings of the A CM Symposium on Theory of Comput-
ing, pp. 1-8.
Peters, S. and R. Ritchie: 1973, "On the Gener at i ve Power of Transformat i onal Gr ammar s' ,
Information Sciences 6, 49-83.
362 ALEXI S MANASTER- RAMER AND MI CHAEL B. KAC
Pike, K. L.: 1954-60, Language in Relation to a Unified Theory of the Structure of Human
Behavior, Summer Institute of Linguistics, Gl endal e, CA. , 3 parts.
Postal, P. M.: 1964a, Constituent Structure: A Study of Contemporary Models of Syntactic
Description, Mout on, The Hague.
Postal, P. M.: 1964b, "Limitations of Phrase Structure Gr ammar s' , in J. A. Fedor and J. J.
Katz (eds.), The Structure of Language: Readings in the Philosophy of Language, pp.
137-51, Prentice-Hall, Englewood Cliffs, NJ.
Pullum, G. K. and G. Gazdar: 1982, ~Natural Languages and Cont ext -fl ee Languages' ,
Linguistics and Philosophy 4, 471-504.
Rfiv6sz, G.: 1976, "A Not e on t he Rel at i on of Turi ng Machines to Phrase Structure Gram-
mars' , Computational Linguistics and Computer Languages 11, 11-16.
Rounds, W. C., A. Manast er-Rarner, and J. Fri edman: 1987, "Finding Nat ural Languages
a Home in Formal Language Theory' , in A. Manast er - Ramer (ed. ), Mathematics of
Language, pp. 349-59, John Benj ami ns, Amst erdam.
Savitch, W. J. , E. Bach, W. Marsh, and G. Safran-Naveh (eds.): 1988, The Formal Com-
plexity of Natural Language, Reidel, Dordrecht .
Uszkoreit, H. and S. Peters: 1986, "On Some Formal Propert i es of Met arul es' , Linguistics
and Philosophy 9, 477-94.
Wasow, T.: 1978, ' On Constraining the Class of Transformat i onal Languages' , Synthese 39,
81-104.
Well, R. S.: 1947, qmmedi at e Const i t uent s' , Language 23, 81-117.
Yngve, V. H. : 1960, "A Model and an Hypothesis for Language St ruct ure' , Proceedings of
the American Philosophical Society 104, 444-46.
Al exi s Manast er- Ramer
Box 704
T. J. Wat son Research Center
I B M Corporat i on
Yorkt own Hei ght s, NY 10598
U. S. A.
Mi chael B. Kac
Depar t ment o f Li ngui st i cs
University o f Mi nnesot a
Mi nneapol i s, MN 55455
U. S. A.

You might also like