It is by no m eans adequate to represent all the criticism s against the sentential
level backw ards pronom inalization rules w ith its extrem e/strong form (“P redictability
R eq u irem en t” (kuno)) under the label o f ‘F orw ard O nly H y p o th esis’ (C arden). There
are, at least, three points o f criticism m ade against the sentence-level backw ards
pro n o m in alizatio n rules, w hich com m only share the follow ing assum ption presented
by R oss (1967).
I w ill assum e that structures underlying sentences like (la) m u st be converted into those that underly (lb) or those th at underly (lc) by a transform ational rule o f PR O N O M IN A L IZ A T IO N .
(I)(a) A fter John A dam s/ w oke up, Jo h n A dam s/ w as hungry. (b) A fter John A dam s/ w oke up, he/ w as hungry.
(c) A fter hez w oke up, John A dam s/ w as hungry.
5-2-1-1. Firstly, as illustrated in R o ss’ passage above, notions o f ‘p ro n o m in alizatio n ’
and ‘backw ard p ro n o m inalization’ necessarily assum e the theoretical postulates o f the
classical generative fram ew ork such as abstract ‘deep stru ctu res’ (la), w hich are
converted to ‘surface structures' (lb and Ic) by the application o f ‘m o v em e n t’ rules
(such as ‘ A d v erb -preposing’ rules) as w ell as by ‘p ro n o m in alizatio n ’ rules. T hese
theoretical postulates are valid only w ithin the generative fram ew ork. A s B olinger
(1977, A bstract) says, “P ronom inalization o f one N P by another is a m isleading
co n c ep t”, those w ho are outside the generative fram ew ork cannot accept the notion o f
5-2-1-2. Secondly, the ‘p ro n o m in a liza tio n ’ rules usually assum e at first the
representation o f a sentence at the abstract deep structure level as in (la) o f R o ss’
presentation, from w hich (lb) and (Ic) are derived. In other w ords, (lb) and (Ic) are
regarded sim ply as different m anifestations o f the sam e content. T his approach is
likely to ignore the functional differences in the actual usage b etw een different
linguistics form s. In effect, it reduces the issue o f the choice o f referrin g expressions
to the m atter o f w hether or not transform ational rules such as A dverb -p rep o sin g are
applied to the sam e abstract deep structure. This kind o f approach is hardly
acceptable for a theoretical p o sition that claim s that different linguistic form s are
necessarily associated w ith the differences in their m eaning, as stated by B olinger
I suggest that w e have been m islead by the concept o f ‘p ro n o m in a liz a tio n ’, o f an N P in one p o sition som ehow controlling an N P in another, and th at w e ought to be asking instead w hat is the basis o f choice for a n o u n or a pronoun in any given position, view ing each as an independent decision.
(B olinger, 1977:4)
5-2-1-3. Thirdly, one o f the m ain criticism s tow ards ‘backw ard p ro n o m in a liza tio n ’ is
th at it is im possible to account for pronom inal references solely on th e basis o f
sentence-level rules; we need to take account o f discourse factors in o rd er to account
for catap h o ra (backw ards anaphora). As Kuno (1975) notes, referring to K arttunen
(1968):
so-called backw ard p ronom inalization involves discourse factors an d is not d escribable in purely syntactic term s. (K uno, 1975:284)
I w ould argue that this criticism tow ards the generative syntactic b ack w ard
p ronom inalization rules is still valid, despite o f C ard en ’s argum ent a g ain st the
C arden (1982:361) says:
W hile there are various problem s in stating a B ackw ards P ronom inalization rule precisely, it is clear that som e such m echanism w ill be needed as long as w e lim it the dom ain o f our rules to a single sentence. B ut w hat happens w hen w e extend our m odel to perm it rules th at have an entire discourse as their dom ain? W e w ill then have a rule o f D iscourse Pronom inalization;
B y asking “w hat happens w hen we extend ou r m odel to p erm it rules that have an
entire discourse as their d om ain”, how ever, C arden actually shares the sam e
theoretical position as K uno, w ith his criticism tow ards the sentential approach to pro
form s, adm itting that backw ards anaphora cannot be handled solely w ith a sentence-
level syntactic approach but has to be accounted for through discourse factors. It
should also be noted that the notion o f 'F irst-m e n tio n ’ itse lf is not a syntactic notion
b u t a discourse notion. A s noted in chapter 4, C arden presents several types o f
genuine backw ards anaphora. A m ong them , it is still possible to argue that the
follow ing four types (a)-(d) can be identified w ithout resort to the influence o f
T able 5.2.1: F our types o f genuine cataphora (C arden, 1982) identifiable by sentential factors.
(a) First M ention: The antecedent is indefinite N P
e.g. W hen she was five years old, a child o f m y acquaintance announced a theory that she w as inhabited by rabbits. {New Y ork T im es, 6 N ov. 78)
(b) S ingular P ronouns B ound by S ingular Q uantifiers.
e.g. ... there are six legally operated and licensed poker cardroom s . . . A s itsm ajor source o f incom e, each club collects a playing fee from the players every h a lf hour. . . {Social P roblem s, 28: 557 (1977))
(c) Plural Pronouns w ith the V ariable R eading.
e.g. D id you know that w hen their w ives leave them , two men in f iv e go bananas? L anniglan's R abbi, TV , fall 1977 (collected by T. D ieterich)
(d) V ariable-R eading P ronouns B ound by G enerics
e.g. N o m atter how innocent he m ay be in his inner soul and in his m otivations, the effective mathem atician is likely to be a pow erful factor in changing the face o f society. (W einer, Ex-Prodigy: 189-190)
T he o ther types o f genuine backw ards anaphora (the F irst-M ention cases involving
definite N P antecedents and the O verriding types), how ever, cannot be identified
w ithout taking account o f discourse factors. W e can tell w h ether the referent o f a
definite N P is F irst-m ention or not only after checking the p receding text. The
o v erriding types, m oreover, necessarily assum e m ore th an one com peting referent