Inheritance and Complementation: A Case 
Study of Easy Adjectives and Related 
Nouns 
Dan Flickinger* 
Hewlett-Packard Laboratories 
John Nerbonne t 
Deutsches Forschungszentrum ffir 
Kfinstliche Intelligenz 
Mechanisms for representing lexically the bulk of syntactic and semantic information for a lan- 
guage have been under active development, as is evident in the recent studies contained in this 
volume. Our study serves to highlight some of the most useful tools available for structured lexical 
representation, in particular (multiple) inheritance, default specification, and lexical rules. It then 
illustrates the value of these mechanisms in illuminating one corner of the lexicon involving an 
unusual kind of complementation among a group of adjectives exemplified by easy. The virtues 
of the structured lexicon are its succinctness and its tendency to highlight significant clusters of 
linguistic properties. From its succinctness follow two practical advantages, namely its ease of 
maintenance and modification. In order to suggest how important these may be practically, we 
extend the analysis of adjectival complementation in several directions. These further illustrate 
how the use of inheritance in lexical representation permits exact and explicit characterizations 
of phenomena in the language under study. We demonstrate how the use of the mechanisms em- 
ployed in the analysis of easy enables us to give a unified account of related phenomena featuring 
nouns such as pleasure, and even the adverbs (adjectival specifiers) too and enough. Along 
the way we motivate some elaborations of the HPSG (head-driven phrase structure grammar) 
framework in which we couch our analysis, and offer several avenues for further study of this 
part of the English lexicon. 
1. Introduction 
The lexicon is a large and complex set of information about the words used in a 
grammar or natural language processing system. Its importance has become more 
central in the research of the past decade, which has seen the rise of radically lexical- 
ized theories such as head-driven phrase structure grammar (HPSG), in which phrase 
structure rules play a vestigial role. Newer theories place increasingly high demands 
on lexical representation. A simple calculation may illustrate the quandary of lexical 
representation: feature systems for contemporary systems normally distinguish at least 
30 features (while 40 or 50 is not rare). The number of values a feature takes ranges 
from 2 to the number of categories (more exactly, to the number of sequences or sets 
of a small size, where all the members of the sequence, etc. are categories). Under 
the undoubtedly optimistic assumption that feature value ranges could be reduced to 
booleans, we still are faced with 230 -=- 109 feature combinations--whose individual 
* 1501 Page Mill Road, Palo Alto, CA 94304-1126, flickinger@hplabs.hp.com. 
t Stuhlsatzenhausweg 3, D-6600 Saarbr/icken 11, Germany, nerbonne@dfki.uni-sb.de. 
(~) 1992 Association for Computational Linguistics 
Computational Linguistics Volume 18, Number 3 
representation is clearly to be avoided, not "solved. ''1 The natural tack is certainly to 
represent just the categories actually used in the vocabulary, but this could incur a 
good deal of redundancy if it meant that each feature combination were represented 
separately on each word. 
The structured or hierarchical lexicon solves this difficulty (cf. Flickinger, Pollard, 
and Wasow 1985 and Flickinger 1987). In structured lexicons, word classes may stand 
in a relationship of inheritance to one another, in which case the properties of the 
bequeathing class accrue automatically to the inheriting class. Once we allow that a 
single class may be heir to more than one bequeathing class, we allow, in principle, that 
no word class property ever need be examined more than once. Thus we eliminate one 
central source of redundancy in lexical specification. One of the goals of this paper 
is to motivate the use of inheritance in lexical specification. To do this, we take a 
narrowly circumscribed phenomenon in English grammar--that of vp-complement- 
taking adjectives, as in hard + to deliver--and spell out the lexical specifications a 
thorough treatment demands. The sheer complexity of these specifications cries out for 
a redundancy-eliminating approach, and we propose a structured lexicon treatment. 
The grammatical analysis not only serves to motivate the general approach, it also 
illustrates several key issues in the design of structured lexicons, such as the use of 
default inheritance, the need for lexical rules, and the range of phenomena amenable 
to this sort of treatment. 
The goals of this paper are to introduce the structured lexicon in a fairly simple 
form, to motivate its basic theoretical device, that of inheritance, with a real example 
taken from an existing system, and finally to show how the elimination of redundancy 
achieved with the structured lexicon aids in maintaining the lexicon. We argue for im- 
proved maintainability by examining concrete extensions and potential modifications 
of the grammatical description provided. We turn now to a brief characterization of 
this phenomenon. 
The rich collection of syntactic and semantic phenomena exhibited by a familiar 
group of adjectives such as tough and easy present a challenge to those who seek to 
provide explicit formal characterizations of linguistic properties. We offer here a de- 
tailed description of the properties of these adjectives, involving optional and obliga- 
tory complementation, control, long-distance dependence, optional modification, and 
specification. The purpose of this description here is not the linguistic analysis itself 
(which we find interesting nonetheless), but rather its use in demonstrating the prac- 
tical utility of inheritance as a tool for linguistic description, and also the predictive 
analytical power that inheritance affords in the study of the lexicon. In illustration of 
the latter, we extend our analysis of easy adjectives to a similar group of nouns such as 
pleasure, and then to the unusual adverbs too and enough, which function as specifiers 
in adjectival gradation. 
The fundamental data are illustrated in (1); examples such as these have not at- 
tracted attention in computational linguistics, even if they have often appeared in 
studies within the generative framework. An early discussion of them is found in 
Miller and Chomsky (1963), with a score and more of additional studies published in 
the years since. Most of the salient properties of these adjectives have already been 
brought to light, but in piecemeal fashion and most often as part of a larger debate 
about the nature of unbounded dependencies, where detailed syntactic and seman- 
tic characterizations of these missing object constructions proved less important. 2 We 
1 Cf. Gazdar et al. 1985, Appendix for a small grammar that nonetheless exceeds the size speculated on 
here. 
2 Related work in theoretical and descriptive linguistics includes Chomsky (1965), Rosenbaum (1967), 
270 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
return to the characteristic properties of these adjectives in Section 3, where they are 
catalogued and given formal representation. 
(1) a. Bill is easy to talk to. 
b. It is easy to talk to Bill. 
c. Bill is easy for Mary to talk to. 
d. It is easy for Mary to talk to Bill. 
We chose this phenomenon as a vehicle to recommend lexical inheritance because 
it illustrates a wide range of grammatical phenomena, aH of which make demands on 
lexical resources (at least in the lexicalized grammar in which the analysis is framed). 
In addition to the grammatical demands, the data justify the use of a lexical rule 
(derivational rule) to relate pairs such as (a) and (b) in (1)--so we shall argue at any 
rate--thus illustrating a further inheritance-like relationship in the lexicon. 
The remainder of the paper is structured as follows: Section 2 summarizes the 
aspects of HPSG that are important to our proposal, and Section 3 develops the fun- 
damental analysis that Section 4 illustrates in a series of analytical "snapshots" of a 
single example. Section 5 suggests extensions of the fundamental analysis, especially 
to further lexical classes (developing the argument that structured lexicons are easily 
maintained and extended), and a final section summarizes and suggests directions for 
future work. Appendix A presents the framework for lexical description developed in 
Flickinger et al. (1985) and Flickinger (1987). The framework is convenient for feature- 
based grammars, but it allows the specification of other lexical properties as well. This 
Appendix presents a notation that is precise while avoiding redundancy, e.g., in char- 
acterizing the kinds of complements that these adjectives permit, and in expressing 
the relationships that hold between pairs like the easy of (la) and that of (lb). Since a 
fundamental claim of hierarchical lexicons is that they eliminate redundancy and thus 
improve modifiability, there is a second appendix, Appendix B, which demonstrates 
the modifiability of the structured lexicon. 
2. Grammatical Theory 
The phenomena involved in the analysis of the easy adjective class illustrate (obligatory 
and optional) subcategorization, control, long-distance dependence, optional modifi- 
cation, and specification (the last in its interaction with adjectival gradation with too 
and enough). As such, it represents an excellent demonstration vehicle for the lexi- 
cal demands of grammatical analysis. Our analysis is formulated within head-driven 
phrase structure grammar (HPSG), the grammatical theory developed by Carl Pol- 
lard and Ivan Sag during the mid and late 1980s. See Pollard (1984; 1985; 1988; 1989) 
and Pollard and Sag (1987; 1988; 1991). As the lengthy list of publications might sug- 
gest, this grammatical theory is well enough documented so that we may restrict our 
Ross (1967), Postal (1971), Bresnan (1971), Chomsky (1973), Lasnik and Fiengo (1974), Jackendoff (1975), 
Chomsky (1977), Fodor (1978), Brame (1979), Nanni (1980), Schachter (1981), Jacobson (1982, 
pp. 221-223), Sag (1982), Maling and Zaenen (1982, pp. 253--254), Kaplan and Bresnan (1982, 
pp. 255-263), Culicover and Wilkins (1984), Jacobson (1984), Gazdar et al. (1985, pp. 150-152), Jacobson 
(1990), Jones (1990), Bayer (1990), and Hukari and Levine (1991). None of these works has attempted a 
thorough descriptive analysis of the range of data we address here, though we are of course indebted 
to these studies for much of the data and many of the generalizations we seek to express. In particular, 
our account is consistent with the brief generalized phrase structure grammar (GPSG) analysis of 
these adjectives given in Gazdar et al. (1985, pp. 150-152) though we embrace a larger range of data 
and extend the analysis to related nouns, a topic rarely discussed since its introduction by Lasnik and 
Fiengo (1974). 
271 
Computational Linguistics Volume 18, Number 3 
remarks here to the distinctive characteristics of the assumptions used here. We assume 
familiarity with feature-based grammars and basic familiarity with HPSG as well. 
In all linguistic theories there is a division of labor between grammatical rules 
and the lexicon, and this concerns the amount of information contained in each. At 
the rule-based extreme lie non-feature-based context-free grammars, where the lexicon 
merely links lexical items to nonterminals; in these grammars it is indeed customary 
to view the lexicon as a set of unary rules. The grammatical rules thus effectively 
encode all linguistic information. At the lexical extreme we find feature-based catego- 
rial grammars, which allow function argument application as the only grammatical 
rule. Here the lexicon bears the burden of encoding linguistic information, and the 
contribution of rules is marginal. We emphasize that HPSG is found very close to the 
lexical extreme, because this highlights the significance of the present work--HPSG is 
a framework whose lexical demands are very nearly maximal. 
Subcategorization information is lexically based in HPSG, much as it is in Cat- 
egorial Grammar (Bach 1988). Grammatical heads specify the syntactic and semantic 
restrictions they impose on their complements and adjuncts. For example, verbs and 
verb phrases bear a feature SUBCAT whose content is a (perhaps ordered) set of fea- 
ture structures representing their unsatisfied subcategorization requirements. Thus the 
feature structures associated with transitive verbs include the information: 
\[subcat: /\[ NPe:acc \]' \[NPcase: nom \]/ \] 
(where hip abbreviates a substantial feature structure.) Applied to adjectival VP com- 
plementation, this treatment of subcategorization leads naturally to the postulation of 
adjectives that subcategorize for VPs, etc. (details follow). 
The significance of subcategorization information is that it represents a (perhaps 
ordered) set of grammatical categories with which a subcategorizer combines in form- 
ing larger phrases. When a subcategorizer combines with a subcategorized element, 
the resultant phrase no longer bears the subcategorization specification--it has been 
discharged. Compare Pollard and Sag (1987, p. 71) for a formulation of the HPSG 
subcategorization principle. 
We shall in general present subcategorization specifications in a slightly different 
way from that above, i.e., not as a single feature whose value is a list, but rather as a 
collection of complement features with category values. Compare Borsley (1987) for 
a development of this approach, which we shall not attempt to justify here. We will 
therefore reorganize the information above in the following way: 
subject: \[ NP case: nom 
°bject: \[ NP \] case: acc 
We choose this representation here only because we find the keywording of gram- 
matical functions, subject, etc., more perspicuous than an encoding in terms of list 
positions, but nothing in the analysis hinges on the one or the other representation. 
We shall furthermore allow that subcategorized elements be either obligatorily sub- 
categorized or optionally subcategorized. Optionally subcategorized elements need 
not be discharged from subcategorization specifications. (This necessitates an obvi- 
ous change to the principle that subcategorization must be satisfied in independent 
272 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
utterances.) In case an element is not discharged, something must be said about its 
semantics. Here we borrow an idea from Situation Theory, and specify that unsatu- 
rated predicate argument structures (or infons; see Devlin 1991) may hold when there 
is some way of filling out the unfilled argument positions so that the result holds. This 
has the effect of existentially quantifying over unfilled argument positions. Linguis- 
tically, there are many other ways in which arguments may be omitted (cf. Fillmore 
1985), but this seems to suffice for the adjectives under examination here. 
Control and modification, the latter being the relation between an adjunct and a 
head, are both lexically realized in the case of the easy adjectives. We regard there as 
being a control relation between for Smith and toget in complex adjectivals such as easy 
for Smith to get (cf. Gazdar et al. 1985: 83ff). Modification plays a role when complex 
adjectivals appear in construction with nominal heads, as in easy job for Smith to get. 
These are common assumptions in the analyses of control and modification. 
Long-distance dependence is treated in HPSG in much the same way it was 
treated in GPSG (cf. Gazdar et al. 1985), and we assume basic familiarity with this 
type of analysis. We recall that the site of a missing element in a "gappy" constituent 
bears a feature SLASH whose value is a specification of the expected material. The 
SLASH specification is propagated by general principles (which we shall not elucidate) 
to the higher level constituents, until it is matched by a "filler" or a subcategorizing 
element. When the gappy constituent is adjoined to a filler or subcategorizing element, 
the result no longer bears the SLASH value. 
Important for our purposes is the possibility of a lexical entry specifying that a 
dependent may contain a gap. (Cf. Gazdar et al. 1985, pp. 150-153 for the first mention 
of this suggestion.) We shall exploit this in the analysis of several word classes below, 
viz., the ones that subcategorize for a VP with an NP in SLASH. It is unusual to find 
a subcategorization specification for SLASH, but not unique: comparatives likewise 
subcategorize for gappy complements, as in seen in examples such as taller than it 
is A wide. We shall require lexical specifications that lead to feature structures of the 
following form: 
stem: easy 
sere: easy( ~\], ^ ~\] ) 
syn.loc.subcat: 
sub j: \[ syn: NP-nom \] 
sem: ~-~ 
pp-for: \[ sem:\[-~ \] 
xcomp: 
syn: VP-inf 
sem: ~\] 
slash: \[ sem:\[~\] \] 
The tag \[~ in the diagram above shows that the semantics of the SLASH value and the 
adjectival subject semantics have been identified. Thus, once a VP/NP has combined 
with this adjective, the semantic contribution of the SLASH element is assumed by 
the subject. Figure 1 shows an analysis tree for an example containing a long-distance 
dependency. 
273 
Computational Linguistics Volume 18, Number 3 
Det 
Y 
NP /\ 
N 
These books 
VP /\ 
V AdjP /\ 
are A.dj 
easy 
VP/NP /\ 
V VP/NP /\ 
to V 
have 
S/NP /\ 
NP VP/NP /\ 
Bob V NP/NP 
read 
Figure 1 
Complex adjectivals such as easy subcategorize for a complement VP containing a "slashed" 
NP, i.e., a VP missing an NP (whose expected position may be arbitrarily deep). 
The variety of linguistic phenomena exemplified in the easy class of adjectives 
guarantees that it is a demanding testing ground for theories of lexical representation. 3 
3. Adjectival VP Complementation 
We assume familiarity with the mechanisms of lexical inheritance and lexical rules in 
the analysis to follow, but we provide an overview of these mechanisms for lexical 
3 It is also worth mentioning that HPSG has also been the subject of intensive implementation activity 
during the past several years; we know of implementations at Hewlett-Packard Laboratories, The 
German AI Center (DFKI), Stanford University, Carnegie Mellon University, The Ohio State University, 
Simon Fraser University, University of Edinburgh, ICOT, University of Stuttgart, the IBM LILOG 
project in Stuttgart, and ATR. We may therefore safely refer the reader to documentations of those 
implementations, even if these are less generally available than the theoretical literature: Proudian and 
Pollard (1985), Nerbonne and Proudian (1987), Franz (1990), Emele and Zajac (1990), and Carpenter, 
Pollard, and Franz (1991). 
274 
Dan Flickinger and John Nerbonne Inheritance and Complernentation 
representation in Appendix A. The fundamental data we shall be concerned with are 
repeated in (2): 
(2) a. 
b. 
C. 
d. 
Other adjectives 
amusing 
annoying 
(3) boring 
comfortable 
confusing 
Bill is easy to talk to. 
It is easy to talk to Bill. 
Bill is easy for Mary to talk to. 
It is easy for Mary to talk to Bill. 
that show this same distribution include the following: 
depressing great nice 
difficult hard painful 
exhausting important tiresome 
fun impossible terrible 
good impressive tough 
Given pairs like (2a,b) and (2c, d), two dusters of properties begin to suggest them- 
selves as part of the definitions of the relevant lexical entries. The first of these clusters 
we will associate with the class of words containing lexical entries for the easy of (2a,c) 
and its counterparts in (3), a class we term SLASH-EASY. The other cluster of prop- 
erties we associate with a second class termed IT-EASY, containing the lexical entries 
for the variant of easy in (2b, d) and its counterparts in (3). We begin by simply iden- 
tifying the relevant properties in each of these two classes, supported by examples as 
necessary; then we provide motivation for factoring these properties into several word 
classes linked by inheritance. 
Adjectives in the IT-EASY class have two obligatory complements, an NP subject 
and a verbal complement; in addition they have one optional complement, a PP headed 
by the preposition for. As seen in (4), the verbal complement can be either infinitival 
or gerundive, and (5) shows that this complement can be a VP even with a PP-for 
present, or an infinitival S, again with or without the optional PP-for complement. 
The subject NP must be the expletive it. 
(4) a. It was great working for Bill. 
b. It was great to work for Bill. 
(5) a. It's easiest for the dogs to feed them at noon. 
b. For the dogs, it's easiest to feed them at noon. 
c. It's easiest for the dogs to be chained up all day. 
d. *For the dogs, it's easiest to be chained up all day. 
e. It's easiest for me for the dogs to be chained up all day. 
f. For me, it's easiest for the dogs to be chained up all day. 
Examples (5e, f) demonstrate that not only VP complementation but also S comple- 
mentation, is involved in easy subcategorization. Note that S complementation never 
requires a controller, and that the PP phrase in such structures is mobile (5f). In addi- 
tion to the conclusion that a variety of complementation schemes are used with easy, 
the data above also demonstrate that the exact specification of the controller (the un- 
derstood subject of the infinitival VP) is nontrivial. Example (5a) demonstrates that the 
PP-FOR complement need not control the VP, and (5b) suggests that noncontrolling 
PPs are more mobile than controllers (5d). 
275 
Computational Linguistics Volume 18, Number 3 
We accommodate these facts semantically by aUowing that easy and similar adjec- 
tives denote two-place relations between individuals and states of affairs. The relation 
holds between the pair, roughly, when it is easy (or convenient) for the individual 
when the state of affairs obtains. Examples (5e,f) show that the individual involved in 
the easy relation need not be involved in the state of affairs, i.e., that there is no nec- 
essary semantic control involved in this relation. 4 The control facts are clear enough: 
when this easy is combined with an S, there is no semantic control; and when it is com- 
bined with a VP, there is no grammatically specified controller of the VP--although 
there may be pragmatic inference about the understood subject. 
Adjectives in the SLASH-EASY class also have two obligatory complements, an 
NP subject and a verbal complement, as well as an optional PP-for complement. In 
contrast to the first class, this class specifies that the subject is a normal (nonexpletive) 
NP, and that the verbal complement must contain an NP gap. Moreover, this verbal 
complement must be infinitival, not gerundive, as seen in (6), and must be a VP, not 
an S, as shown in (7). 5 
(6) a. Bill was great to work for. 
b. *Bill was great working for. 
(7) a. For me, Bill was easy to talk to. 
b. *Bill was easy for me for Mary to talk to. 
In the word class hierarchy we assume, sketched in Appendix A, there is a word 
class CONTROL, which introduces a verbal complement subcategorization, and which 
serves as the superclass from which both of the classes IT-EASY and SLASH-EASY 
inherit. However, neither of these classes is an immediate subclass of CONTROL; we 
draw on the data provided in (8) and (9) below to motivate two intermediate word 
classes that will stand between CONTROL and these two in the hierarchy. 
The English lexicon contains two more groups of adjectives that have much in 
common with the two variants of easy introduced above, but must be kept distinct. 
Lasnik and Fiengo (1974:535) identified a set of adjectives including pretty and melodi- 
ous, illustrated in (8). 
(8) a. Disneyland is pretty to look at. 
b. Sonatas are melodious to listen to. 
c. *It is pretty to look at Disneyland. 
d. *It is melodious to listen to sonatas. 
e. ?Disneyland is pretty for children to look at. 
f. ?Sonatas are melodious for serious musicians to listen to. 
4 There is an interesting pragmatic problem lurking in the control specifications involved here. If one 
specifies the control relationships exactly, then one needs to postulate systematic structural ambi- 
guity in examples such as (5c), where the sequence of PP and VP may or may not be analyzed as an 
constituent. This seems plausible, but then we would like to have a pragmatic account of why there is 
normally no distinction, i.e., why the control relationship is inferred, or, equivalently for all intents and 
purposes, why the S reading is so strongly preferred. 
5 Hukari and Levine (1991) note in passing that there is a group of closely related adjectives such as 
worth that do take a gerundive complement instead of the usual infinitival complement, as in That 
article is not worth looking at. The extension of our analysis to worth is straightforward, but not given here. 
276 
Dan Flickinger and John Nerbonne Inheritance and Complernentation 
Members of this class of adjectives share much in common with the SLASH-EASY 
adjectives, but have two significant differences: first, as shown by (8c,d), they do not 
have a corresponding entry with an expletive it subject, and second, they assign a real 
thematic role to their subjects. That is, (8a) entails that Disneyland is pretty, while (la) 
does not entail that Bill is easy. The two-place relation suggested above for IT-EASY 
and SLASH-EASY adjectives could not account for the validity of this inference, since 
the subject of the adjective plays no direct role in the relation whatsoever. A distinct 
semantic relation is called for here, one in which the subject does play a role (which 
effectively makes this class a kind of EQUI adjective in contrast to the raising easy). 
It also appears that these adjectives do not permit the optional PP-for complement 
licensed by easy in (lc), though judgments are less clear. In order to express these 
differences, we introduce a class SLASH-COME which will include the entries for 
pretty adjectives, and which will also serve as the class from which SLASH-EASY 
inherits. 6 
Similarly, English has a set of adjectives that have much in common with the 
IT-EASY adjectives of (lb, d), but with no counterparts of the SLASH-EASY type. 
(9) a. It is possible to talk to Bill only at breakfast. 
b. It is unnecessary to fire Bill. 
c. *Bill is possible to talk to only at breakfast. 
d. *Bill is unnecessary to fire. 
The second principal difference between adjectives such as possible and those of the 
IT-EASY class is that the former do not permit an optional PP-for phrase complement; 
they do allow the verbal complement to be either a VP or an S (containing a PP-for 
subject), but (10) shows that if a PP-for is present, it must be contained within the 
complement. 
(lO) a. It is unnecessary for Mary to fire Bill. (M firing B) 
b. *For Mary, it is unnecessary to fire Bill. (M firing B) 
c. *It is unnecessary for Mary for you to fire Bill. 
Again, we express the distinction between the set of adjectives like possible and the 
IT-EASY adjectives by introducing a fourth class IT-SUBJ parallel to SLASH-COME 7 
These four class definitions, together with one supporting class, are given in (11- 
16), with the Superclasses attribute showing the relevant inheritance relations. 
(11) 
IT-SUN 
Superclasses 
Complements 
Subject-Features 
Subject-Role 
XComp-features 
Control 
(NForm it) 
none 
(VForm Infinitival) (Complete + -) 
6 Other adjectives of this SLASH-COMP class include delicious, handsome, attractive, and lovely. 
7 Additional members of this IT-SUBJ class include essential, necessary, sad, silly, and illegal. 
277 
Computational Linguistics Volume 18, Number 3 
The disjunctive specification (Complete + -) overrides the default (Complete -) 
specified in the CONTROL class, and means that the verbal complement may be either 
a VP (Complete -) or an S (Complete +). This is an example of further specifying a 
default specification. 
(12) 
SLASH-COMP 
Superclasses Control 
Complements 
XComp-Subj-Semantics 
XComp-features 
x 
(SLASH (Category Noun) 
(NForm Normal) 
(Complete +) 
(Predicative -) 
(Case Accusative) ) 
(Semantics 
Subject-Semantics) ) 
The SLASH feature on the XComp specifies that the VP must contain a gap for 
a normal (non-expletive) noun phrase, which is accusative case and which is not 
predicative. This nonpredicative specification serves to exclude examples like *Bill is 
difficult to become (assuming the complement of become is predicative), since the gap 
for that complement would fail to satisfy the restriction on SLASH given in (12). The 
SLASH specification furthermore notes that the SLASH semantic value is identical 
to that of Subject-Semantics. As was explained in Section 2 above, this is the form a 
lexical specification of semantic coindexing takes. 
The controller of the controlled complement is specified through the attribute 
XComp-Subj-Semantics; for example, in CONTROL, this attribute has the value Subject- 
Semantics, since subjects are default controllers. But the complements of SLASH- 
COMP are not grammatically controlled (cf. (8e,f)), a fact that requires an overwriting 
specification. The semantic variable x is used here because it will not represent the 
semantics of any grammatical complement, which ensures that no grammatical con- 
trol is effected (see examples (9a,b)). This is an example of a subregularity appearing 
within an exceptional specification. 
The classes for the two variants of easy adjectives we have discussed have one 
cluster of properties in common: they both license the optional PP-for phrase seen in 
preceding examples. To further reduce redundancy, we define in (13) the class FOR- 
EXPERIENCER, from which the two classes in (14-15) also inherit. 
(13) 
FOR-EXPERIENCER 
Superclasses 
Complements 
PP-for-Features 
PP-for-Oblig 
PP-for-Role 
PP-for 
(Category Preposition) (Lexical -) 
(PForm For) 
No 
For 
278 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
(14) 
IT-EASY 
Superclasses It-Subj, For-Experiencer 
Complements 
XComp-Features (VForm Infinitival Gerund) 
SLASH-EASY 
Superclasses Slash-Comp, For-Experiencer 
Complements 
Subject-Role none 
XComp-Subj-Semantics PP-For-Semantics 
(15) 
As expected, the IT-EASY class eases one restriction on the verbal complement; 
note too that no controller is specified, in keeping with remarks on (5). On the other 
hand, the SLASH-EASY class blocks inheritance of the subject's thematic role assign- 
ment (the default value having been specified in the INCOMPLETE class from which 
CONTROL inherits), and alters the control relationship (inherited from SLASH-COMP 
and ultimately from CONTROL) so that the PP-For phrase rather than the subject of 
easy is interpreted as the subject of the VP complement. These are two further examples 
of the way in which default overwriting is employed; note that the latter represents a 
subregularity within a subregularity (cf. SLASH-COMP). 
With reasonable assumptions about the definitions of other relevant classes in the 
hierarchy, along with an explicit definition of the class ADJECTIVE, provided here for 
clarity in (16-17), we can introduce the (sparse) lexical entries for the two variants of 
easy employed in (la,b), as given in (17,18): 
(16) 
ADJECTIVE 
Superclasses Major 
Features (Category Adjective) (Predicative + -) 
(17) 
easy-la 
Superclasses Adjective, Slash-Easy 
Semantics easy 
Spelling "easy" 
Phonology /izi/ 
easy-lb 
Superclasses Adjective, It-Easy 
Semantics easy 
Spelling "easy" 
Phonology /izi/ 
(18) 
279 
Computational Linguistics Volume 18, Number 3 
FOR-EXPERIENCER 
CONTROL 
ITq IUBJ SLAStt-COMP 
IT-EASY SLASH-EASY 
easy-lb easy-la 
ADJECTIVE 
Figure 2 
The structure of word classes directly involved in the definition of complex adjectival lexical 
entries. 
Pairs of sparse lexical entries like those in (17,18) are related by a lexical rule 
we label LR-EASY, which simply states that for each member of the class IT-EASY 
there exists a corresponding lexical entry belonging to the class SLASH-EASY, with 
everything but the Superclasses property identical in the two (sparse) entries. 
Rule 
LR-EASY lexical rule 
LR-EASY 
LE2-Classes - IT-EASY = LE1-Classes - SLASH-EASY 
Once each of (17) and (18) are fleshed out to include all of their inherited properties, 
they will of course be quite distinct, as needed to ensure the differences in distribution 
that we have described. Figure 2 summarizes the inheritance relationships thus far. 
4. An Example Analysis 
The purpose of this section is primarily illustrative--we would like to demonstrate the 
effect of the lexical specifications suggested on more familiar elements of grammatical 
analysis, viz. phrases, parse trees, and predicate logic representations. 
The semantics of the easy-SLASH construction, which treats easy as a relation 
between an individual and a state of affairs, is treated as a normal case of lexically in- 
280 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
herited semantics, i.e., one in which the relation denoted has an argument place for the 
denotations of each of the role-playing complements, in this case the PP-FOR phrase 
and the XCOMP. This class of adjectives also has a SUBJECT among its complements, 
but it bears no role (as word class SLASH-EASY specifies), because this is a raising 
construction. For this reason, there is no argument place reserved in the semantics of 
easy-SLASH adjectives for the subject's denotation. To conserve space in the diagrams 
below, relations will be specified not using the keyword coding shown in word class 
and lexical entry specifications (above), but rather in the more familiar order coding. 
In order to make not only the semantics but also the syntax somewhat clearer 
in its intended effect, we include here somewhat elaborate analytical sketches of the 
complex adjectival phrase easy to get Mary to hire in (19): 
(19) Tom is easy to get Mary to hire. 
To begin, we note that the sparse lexical entry for the SLASH-EASY version of easy 
may be filled out to a much richer structure if inherited properties are noted explicitly. 
easy-la 
Features 
Complements 
PP-for-Features 
PP-for-Oblig 
PP-for-Role 
PP-for-Semantics 
XComp-Features 
XComp-Subj-Semantics 
XComp-Oblig 
XComp-Semantics 
XComp-Role 
Subject-Role 
Semantics 
Spelling 
Phonology 
(Category Adjective) (Predicative + -) 
PP-for, Subject, XComp 
(Category Preposition) (Lexical -) 
(PForm For) 
No 
For 
PP-For-Semantics 
(Category Verb) (Complete -)(Lexical -) 
(SLASH (Category Noun) (Complete +) 
(NForm Normal) (Predicative -) 
(Case Accusative) ) 
(Semantics Subject-Semantics) ) 
PP-For-Semantics 
Yes 
XComp-Semantics 
State-of-Affairs 
none 
easy 
"easy" 
/izi/ 
The features noted above were specified by the lexical entry together with the classes 
ADJECTIVE, SLASH-EASY, SLASH-COMP, FOR-EXPERIENCER, and CONTROL. Fur- 
ther subject properties would be inherited from INCOMPLETE, but for brevity these 
are not listed. (Of course many other properties, including e.g., gradation properties 
and the applicability of lexical rules, have likewise been suppressed in the interest of 
clarity in presentation.) This lexical description translates fairly directly (with some 
further simplifications and abbreviations) into a feature structure of the sort used by 
HPSG grammars. 
281 
Computational Linguistics Volume 18, Number 3 
stem: easy 
syn.loc.head: Adj 
sem: easy( \[~,^ \[~) 
syn.loc.subcat: 
sub j: \[ syn: NP-nom \] 
sem: ~\] 
pp_for: \[ sY n: PP-f°r \] 
sem: \[\] 
syn: VP-inf 
sere: ~\] 
xcomp: subject.sem: \[~ 
slash: \[ 2eYnm:~ -acc \] 
We would like to draw attention to two semantic coindexings in the structure, 
which are lexically specified and which simplify subsequent (grammatical) processing. 
The coindexing of the xcomp's subject with the pp-for is effected in the SLASH-EASY 
word class, and the semantic coindexing seen above is just a consequence of that. The 
coindexing of the xcomp's slash's semantics value with the subject's semantics, on the 
other hand, derives ultimately from SLASH-COMP. 
In Figure 3 we examine the combination of a token from this class of easy adjectives 
and a VP/NP. The very sparse specification of the mother phrase's features is, in fact, 
solely for purposes of legibility--all of the information specified on the mother node 
may be derived from general HPSG principles, so that nothing is specified, e.g., on the 
rule that licenses head-complement combinations. The fact that the semantics attribute 
is identified with the subcategorizer's semantics follows from the HPSG Semantics 
Principle, which states that the semantics of a phrasal node is always to be identified 
with the semantics of a head in a head complement combination. The fact that the 
slash value of the mother structure is empty follows from the Binding Inheritance 
Principle, which states that slash values are collected going up a tree--unless a head 
subcategorizes for an element containing a slash value, in which case the slash satisfies 
the subcategorization requirement. The identification of the feature structure labeled 
V~, which is just the representation of the phrasal node dominating to get Mary to 
hire, with one of the adjective's subcategorization specifications, that labeled ~\], is 
just a condition for the applicability of the head-complement rule, not an additional 
specification. Of course, the phrasal node is massively underspecified here, but the 
suppressed information is predictable, not merely hidden. 
This is an intriguing aspect of HPSG, but we dwell on it here for self-serving 
purposes. If the properties of the phrasal combination of this fairly intricate syntactic 
structure require no further comment, that is largely because the lexicon has provided 
a wealth of richly structured representation. This would hardly be feasible in the 
absence of efficient and sophisticated lexical representation mechanisms. 
282 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
Adj 
easy 
slash: 0 
AdjP 
to get Mary to hire t 
• stem: easy 
sem: \[~ easy( \[-i-\], ^ ~\] ) 
s bj: \[som\[\] \] 
pp or: \[ sem:\[  \] 
comps: subj: \[ som:\[  \] 
xcomp:\[\] sem:\[\] 
L slash.sem: \[\] 
sere: get(x,m, 1 \[\] " hire(m, ~\])) 
slash.sere: \[\] 
Figure 3 
The combination of complex adjective and slashed VP complement. 
To complete this illustration, we spell out the effects of unification on the structure 
above in Figure 4. Note in particular that because the slash semantics on the VP phrase 
is identified with the slash semantics on the subcategorized-for VP, which in turn is 
identified with the semantics of the subject for easy, the resultant phrase will bind its 
subject to the deeply embedded object argument position of the verb hire. This takes 
place even though the subject plays no role in the easy relation itself. This is exactly 
what is wanted semantically of a raising construction. 
5. Extensions and Lexical Maintenance 
The structured lexicon aims ideally at a redundancy-free specification of all lexical 
properties, and indeed, it achieves this largely through the use of inheritance. While 
we do see scientific parsimony as an end in itself, we see two further advantages in 
the employment of the structured lexicon, one scientific and one practical. The scien- 
tific advantage of the structured lexicon is that it identifies significant classes in the 
language. In a feature system with approximately 30 atomic features (including seman- 
tics), each of which ranges over approximately 10 values, it is certainly striking that 
we never see the need to distinguish 1030 classes of items. In fact we distinguish ap- 
proximately 300 lexical classes in HP-NL, a large system with very broad grammatical 
coverage (see Nerbonne and Proudian 1987). 
283 
Computational Linguistics Volume 18, Number 3 
stem: easy 
sem.logic: \[\] easy( z, ^ get (z, m, 
" hire(m, T,~\] U ~\]) ) 
comps:\[ subj: \[sem:\[~\]U\[~\] \] \] 
Aa5 
to get Mary to hire t 
Figure 4 
The result of combining complex adjective and slashed VP complement. Note that the subject 
of easy is still semantically coindexed with the missing VP object. 
But the practical advantage of the structured lexicon may ultimately also be of 
scientific value, and that is because a structured lexicon is more easily maintained 
and extended than a nonstructured one. This advantage derives immediately from the 
characteristic that lexical properties are normally specified only once. Modifications 
tend then to be minimal and extensions less frightening. The ultimate scientific benefit 
this may bring derives from the fact that it is then easier in systems with structured 
lexicons to experiment with grammatical description. 
The following section is an attempt to buttress the claim that structured lexicons 
are easily extended. We examine therefore extensions to the analysis above of adjec- 
tives that govern VP complements--to nouns with similar subcategorizations, to the 
adjectival specifiers too and enough, and to adjectives that govern S complements rather 
than VP complements. 
5.1 Pleasure Nouns 
Adjectives like easy have been the most widely studied group of lexical types that 
populate the classes introduced in the analysis above, but they do not have exclusive 
claim to those classes. Lasnik and Fiengo (1974) observed that the English lexicon also 
contains a group of nouns with similar properties, as illustrated in (20-21), 
(20) a. Nureyev is a pleasure to watch. 
b. This course is a breeze to pass. 
c. Venice is a delight to visit. 
(21) a. It is a pleasure to watch Nureyev. 
b. It is a breeze to pass this course. 
c. It is a delight to visit Venice. 
Like the adjectives discussed above, nouns such as pleasure have two variants, one 
that appears with an ordinary NP subject and an infinitival complement containing an 
284 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
NP gap; and one that selects an expletive it subject and an infinitival complement with 
no gap. Given the word class definitions developed on the strength of the adjectival 
examples, an obvious analysis of the nominal examples suggests itself: pleasure, like 
pleasant, has one lexical entry belonging to the SLASH-EASY class, and a second entry 
that inherits from the IT-EASY class. The (sparse) descriptions of both entries are given 
in (22-23), parallel to those for easy given in (17-18) above, the salient difference being 
that the noun entries inherit from the class Common-Noun where the adjective entries 
inherited from the Adjective class. 8 
(22) 
pleasure-la 
Superclasses Common-Noun, Slash-Easy 
Spelling "pleasure" 
Semantics.Pred pleasure 
Phonology /plEzhr/ 
pleasure-lb 
Superclasses Common-Noun, It-Easy 
Spelling "pleasure" 
Semantics.Pred pleasure 
Phonology /plEzhr/ 
(23) 
Having declared nouns like pleasure to have entries that are members of SLASH- 
EASY and IT-EASY, nothing more needs to be said in order to capture the syntactic 
relationship between these two forms of pleasure. The lexical rule we proposed earlier 
to link pairs of adjectives like the two variants of easy is defined as a regularity holding 
between the two classes SLASH-EASY and IT-EASY, making no mention of the class 
ADJECTIVE in its formulation. Hence it also serves to link the pair of noun entries in 
(22-23). 
Some further explanation needs to be provided about the semantics of this class 
of nouns, since the nouns do seem semantically anomalous even if we shall maintain 
that all of the apparent anomaly ultimately stems from their having a subject--and 
thus being available for control (by be and other raising verbs). In general a common 
noun is interpreted as a relation between a theme argument and the denotation of its 
complements, if there are any. For example, friend is interpreted as a relation between a 
theme argument and the denotation of the complement PP-OF phrase. We refer to the 
theme argument of the relation denoted by the common noun as its denotation. An 
apparent peculiarity of nouns such as pleasure is that there appears to be no denotation 
of the noun in the usual sense, e.g., in (20a). At issue is whether there is any theme 
argument position for the "pleasure" in the relation denoted by pleasure. That is, does 
pleasure denote the same two-place relation between individuals and states of affairs 
that pleasant does, or is there a third argument position in pleasure that is occupied by 
an (abstract) "pleasure" individual? 
The suspicion that no denotation is involved likely stems from our intuition that 
we do not seem to refer to an object that is a pleasure in uttering either (20a) or (21a), 
at least not any more than we would if we had used pleasant in the place of a pleasure. 
8 Other nouns in this class include disappointment, ordeal, challenge, joy, inspiration, and privilege. 
285 
Computational Linguistics Volume 18, Number 3 
Now this suggests that the noun (phrase) is used predicatively, much as many noun 
phrases are after the verb be. Compare Tom is a linguist. 
This does not help a great deal, however. Even though the analysis of predicative 
NPs is an old topic semantically (cf. the definition of be in Montague 1973, p. 261), 
there has been essentially no successful attempt to treat predicative nouns as if they 
had no denotation. Any attempt to do so seems to run afoul of the standard (if limited) 
determination and adjectival modification found in phrases such as no great pleasure to 
watch; at least such examples point out the inevitable duplication a semantic analysis 
would incur if predicative nominals had no denotation. 
We therefore interpret pleasure as a three-place relation 
pleasure (theme : e~ for : x~ soa : s) 
which obtains just in case e is the pleasure x has in case s. It should of course also 
turn out that this relation for some e holds iff pleasant (for : x~ soa : y), but we will 
not be concerned with showing that here. e provides a denotation that is subject to 
determination (no) and (intersective) adjectival modification (great). Under this anal- 
ysis, a pleasure to watch and no pleasure to watch denote quantifiers, i.e., in each case 
a set of properties of pleasures (e's from above). Of course, a quantifier does not by 
itself represent a proposition, something that could be true or false---for that it must 
be paired with a property. In these cases, the relevant property is always the universal 
(existence) property; i.e., utterances of sentences such as (20a) are true just in case there 
is a pleasure of the relevant kind (and mutatis mutandis for the negative existentials). 
We therefore postulate that the predicate be in these sentences denotes the universal 
property. 9 
What is striking about this proposal is that it assigns the common noun pleasure 
exactly the semantics the general scheme predicts--a relation between a theme and 
the denotations of other complements. For this reason, the word classes for pleasure 
nouns make no special stipulations about semantics. 
We therefore derive feature structures such as the following, which are used in the 
syntax and semantics processing of the word pleasure. The first structure represents 
the member of the SLASH-EASY class, and the second the member of the IT-EASY 
class. (We have simplified the structures to highlight the semantically relevant parts.) 
stem: "pleasure" 
sem.logic: pleasure (e, ~-~ \[\] ) 
subcat: 
subj: I sere:\[\] \] 
pp-for: I sem:\[\] \] 
\[ subj: \[\[sem: \[\] \]1 
xcomp: logic: 2~ 
sem: slash: 
9 In fact, we do not stipulate a peculiar semantics for the raising verbs (such as be) that are involved 
here. Instead, we allow be to denote the identity relation, which holds of a single argument just in case 
there is some way of filling in the missing argument--i.e., in case the first exists. This follows from the 
general treatment of unsaturated relations in Situation Theory (cf. Section 2 under subcategorization). 
Note, however, the one exceptional aspect, i.e., that the subject of the verb be is not linked to any 
argument position in the relation denoted by the controlled complement (in this case, pleasure). 
286 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
stem: "pleasure" 
sere.logic: pleasure (e, ~-~,^ ~-\] ) 
comps: 
subj: NP-it 
pp-for: \[ sem: \[~ \[\] 
xcomp:Isubject: sem:~\] \] \] 
sere: \[~ 
On the other hand, the noun classes are exceptional in that the nouns involved 
have subjects--a property they inherent finally from INCOMPLETE, in the one case 
through CONTROL, IT-SUBJ, and IT-EASY; and in the other from CONTROL, SLASH- 
COMP, and SLASH-EASY. It is this property, shared by the NPs they give rise to, 
that explains (i) their ability to be controlled, e.g., by the verb be--only unsaturated 
phrases are subject to control; (ii) their inability to function in normal NPs, e.g., in the 
subject position of any intransitive verb; and finally (iii) the fact that they can stand in 
construction with the main verb be without being asserted to be identical to its subject. 
We turn now to further points on the syntax of the pleasure nouns. The two defini- 
tions of entries for "pleasure" also predict the grammaticality judgments seen in (24), 
analogous to the examples given above for adjectives, and based on the definitions 
given for the IT-EASY and SLASH-EASY word classes. 1° 
(24) a. Nureyev is a pleasure for us to watch. 
b. It is a pleasure for us to watch Nureyev. 
c. For us, Nureyev is a real pleasure to watch. 
d. *For us, Nureyev is a real pleasure for our parents to watch. 
e. For us, it is a real pleasure for our parents to watch Nureyev. 
f. It is a real pleasure for us for our parents to watch Nureyev. 
g. *Nureyev is a pleasure watching. 
h. It is a pleasure watching Nureyev. 
10 Nothing we have said so far captures the fact that some pairs of members of these two classes, such as 
"pleasant" and "pleasure," are morphologically related. We do not offer here a proposal for capturing 
nonproductive regularities of this kind, though some extension of the lexical rule mechanism might 
serve, an extension that would depend heavily on the ability to specify negative exceptions to lexical 
rules, given examples like the following. 
(i) It is difficult to hire Bill. 
(ii) *It is a difficulty to hire Bill. 
(iii) *Bill is a difficulty to hire. 
(iv) It is impossible to work with Bill. 
(v) *It is an impossibility to work with Bill 
(vi) *Bill is an impossibility to work with. 
287 
Computational Linguistics Volume 18, Number 3 
Recalling further that the adjectives we looked at above fell into not two but four 
distinct classes, we might expect to find nouns as well that belong to the other two 
classes, IT-SUBJ and SLASH-COMP. Such instances are found in English, as illustrated 
for IT-SUBJ nouns by the examples in (25), and for SLASH-COMP nouns by those in 
(26), drawn from Lasnik and Fiengo. 11 
(25) a. It would be a mistake to fire Bill. 
b. It was a shock to find Bill here. 
c. *Bill would be a mistake to fire. 
d. *Bill was a shock to find here. 
(26) a. This room is a pigsty to behold. 
b. Nureyev is a marvel to watch. 
c. *It is a pigsty to behold this room. 
d. *It is a marvel to watch Nureyev. 
The noun mistake and the adjective possible have in common just those properties 
specified by the IT-SUBJ class (together with its superclasses); and like the differences 
between pleasure and easy, their differences result from mistake being a member of the 
COMMON-NOUN class while possible inherits from the ADJECTIVE class. Since the 
lexical rule relating the two variants of pleasure (and the two variants of easy) is defined 
to link members of the two classes SLASH-EASY and IT-EASY, the rule correctly does 
not predict the existence of similar alternate entries for nouns like mistake and pigsty. 
Interaction with lexical rules. Given that the domain of lexical rules is always one 
or more word classes, and that the LR-Intraposition rule is defined on the IT-SUBJ 
class, we predict the grammaticality of the following examples with pleasure nouns, 
since they also have entries belonging to the IT-SUBJ class, and should be expected to 
conform to the LR-Intraposition rule. Here again, the combined devices of inheritance 
and lexical rule produce the desired results for nouns without requiring that anything 
be added to the analysis motivated from data on adjectives and verbs. 
(27) a. (For me) to stay another day would be a real pleasure. 
b. It would be a real pleasure (for me) to stay another day. 
c. To visit Venice now might be a disappointment for you. 
d. It might be a disappointment for you to visit Venice now. 
5.2 Too and Enough 
To drive home our central point about the expressive and predictive power of inher- 
itance in lexical representation, we turn to a third, small class of lexical entries that 
show complementation properties like those we have already seen. Jackendoff (1972) 
noticed that the words too and enough also appear in constructions with an infinitival 
complement that contains an NP gap, as illustrated in (28) with examples drawn from 
11 Additional IT-SUBI nouns include battle, disgrace, error, honor, relief, shock, and surprise. Other 
SLASH-COMP nouns include beauty and terror. 
288 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
Lasnik and Fiengo (1974). 12 
(28) a. The mattress is thin. 
b. *The mattress is thin to sleep on. 
c. The mattress is too thin to sleep on. 
d. The football is soft. 
f. *The football is soft to kick. 
g. The football is soft enough to kick. 
In particular, the examples in (29) suggest that these adverbs select for comple- 
ments that are the same as adjectives like pretty, entries that are not related via lexical 
rule to variants that license an expletive it subject. 
(29) a. *It is too thin to sleep on this mattress. 
b. *It is soft enough to kick this football: 
Informally, it seems that when too or enough combines with an ordinary adjec- 
tive, the resulting phrase (too thin and soft enough) exhibit complementation properties 
very much like those of pretty adjectives. By defining the lexical entries for these two 
adverbial specifiers as members of the SLASH-COMP class, we begin to provide an 
account for examples (28c,g) as well as those in (29). The entry for too is given in (30), 
inheriting both from the ADVERB class and from the SLASH-COMP class; the entry 
for enough is similar, leaving out of the present discussion an account of the linear 
order difference between the two adverbs with respect to the adjective they modify. 
(30) 
"too" 
Superclasses Adverb, Slash-Comp 
Spelling "too" 
Phonology /tu/ 
With the inclusion of this class of adverbs, our lexical subhierarchy involving com- 
plementation of slashed VPs has grown to a point where it surely demonstrates the 
virtues of the structured lexicon approach. Figure 5 illustrates the more complete struc- 
ture. It is a curious fact that the number of lexical classes does not grow enormously 
even while fairly detailed analyses involving very different grammatical areas are un- 
dertaken. In several years of development at Hewlett-Packard Laboratories involving 
detailed analyses of dozens of constructions, the number of word classes never ex- 
ceeded 400. This must be due finally, not to the lexical analysis tool, but rather to the 
tendency of language to reuse significant classes. 
This analysis of these two unusual adverbs has left begging an important is- 
sue about how the complementation specifications provided by too are propagated 
up to the phrase too thin. 13 We have said little here about how lexically supplied 
12 Baltin (1987) presents a more recent analysis of these "degree complements." 13 One might be tempted to try a lexical rule approach that would treat 
too thin as a derived lexical item that selects for a VP complement. But slightly more complicated examples quickly render this approach 
untenable. Cf. This country is too thinly populated to worry about (where we take the scope of the specifier 
too to be thinly populated). Here, the lexicalized form that selects for a VP complement would have to be 
289 
Computational Linguistics Volume 18, Number 3 
CONTROL 
IT-SUBJ SLASH-COMP 
FOR-EXPERIENCER 
IT-EASY SLASH-EASY 
/~~ ADJECTIVE 
pleasure-la ~ j/ ~~ 
pleasure-lb ~\ 
too 
Figure 5 
The lexical subhierarchy involving elements that govern "slashed" verb phrases. Note that the 
original hierarchy needed very little modification, merely addition. We speculate that this is 
due to the fact that significant classes are being identified in detailed grammatical description. 
There is also a version of too that inherits from ADVERB and IT-EASY that is not shown (since 
it was not discussed). The asymmetry is only apparent. 
subcategorization information is employed in parsing, referring the reader to full ac- 
counts given in Pollard and Sag (1987) and related references. Yet it is clear that some- 
thing more must be said about this construction, given that in HPSG it is the syntactic 
head of a phrase that imposes constraints on its complements; and we assume that 
thin, not too, is the head of the phrase too thin to sleep on. To motivate the necessary 
elaboration of our analysis for these two adverbs, we turn to one more set of data 
involving gappy infinitival complements, one that has received little study to date. 
Excursus on subcategorization transfer. As the example in (31) shows, adjectives such 
as easy appear not only in predicative constructions like those illustrated above, but 
also as nominal modifiers. 
(31) John is an easy man to talk to. 
too thinly populated, a consequence we regard as unacceptable. 
290 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
While the example in (31) is good, employing the easy that belongs to the EASY- 
SLASH class, the examples in (32-33) are ungrammatical. The analysis we have pro- 
vided thus far does not yet explain the grammaticality of (31) and the ungrammati- 
cality of (32,33). 
(32) a. *John is an easy to talk to man. 
b. *John is an easy man. 
(33) a. *John is an easy man to talk to Bill. 
b. *John is a man easy to talk to Bill. 
We will focus on explaining the grammaticality of (31), assuming that the right 
syntactic structure for the sentence is the binary-branching structure given in (34), 
where easy forms a constituent with man, and where to talk to is sister to the phrase 
easy man. We adopt the binary structure largely because it will simplify the exposition 
here; it might be equally defensible to hold that easy, man, and to talk to are all sisters 
of a single phrase24 
What is awkward about this structure is that the head noun man does not by itself 
subcategorize for the VP/NP. ~5 Rather, it seems that when easy combines with man, the 
resulting phrase has a subcategorization list that contains not only the optional and 
obligatory complements that man started out with, but also the obligatory VP/NP com- 
plement and the optional For-PP controller required by easy. No mechanism presented 
so far provides for an adjunct combining with its head to affect the subcategorization 
of that head or of the resulting phrase. Yet if the phrase structure proposed in (33) is 
correct, some kind of merging of subcat information between adjunct and head must 
be provided for. 16 
14 And it is worth noting that the alternative constituent structure would not modify the head 
relationship, and therefore would not substantially alter the analytic problem--that of explaining how 
a complement to talk to can be licensed by a nonhead. 
15 At least not with the intended reading. There is a suspiciously similar construction, illustrated in (i), 
which might be expected to shed some light on the proper analysis of (31), but which has a restricted 
enough interpretation to suggest that it should be treated separately, probably inheriting a specification 
from the more general construction exhibited in (31). 
a. John is a man to admire. 
(i) b. Mary is a woman to emulate. 
c. This is a word to keep on the tip of your tongue. 
These examples seem to mean something like John is a good man to admire or Mary is a good woman to 
emulate, where the semantic contribution of good has been incorporated into the N-VP/NP construction 
in (i). To test this, consider the examples in (ii), where the good reading should lead to an anomalous 
interpretation, and does (cf. the corresponding examples in (iii)). 
(ii) a. ?Mary is a person to underestimate. 
b. ?Sharks are animals to tame. 
(iii) a. Mary is an easy person to underestimate. 
b. Sharks are difficult animals to tame. 
Given the constrained interpretations of examples like those in (i-ii), it does not seem defensible to 
treat easy man to talk to as simply the modifier easy combining with man to talk to. Any such attempt 
would be strained in accounting for (ii); in addition, such an analysis would leave unexplained the 
ungrammaticality of *John is an easy man. 
16 It is probably worth noting that extraposition seems unlikely to be generalizable to all cases involving 
transferable subcats, at least if extraposition is to be bounded uniformly: 
(i) An easy man to talk to arrived yesterday. 
(ii) *An easy man arrived yesterday to talk to. 
291 
Computational Linguistics Volume 18, Number 3 
S /\ 
NP VP /\ 
Tom V NP /\ 
Is Det N / 
an N /\ 
A~ N 
easy man ~o 
\ 
VP/NP /\ 
V VP/NP /\ 
V PP/NP /\ 
talk P NP/NP 
~o t 
Figure 6 
Complex adjectivals "wrapped" around a modified noun. Note that the N head of the 
I~I constituent in construction with the complex adjectival has not licensed it. 
Subcategorization transfer has taken place. 
The examples in (34-35) illustrate that the flow of information from an adjunct's 
list of subcats to the head's must be quite restricted; it would not do to simply merge 
the Complements list of any adjunct with that of the head in every case. 
(34) a. *an eager man to please 
b. *a fearful man of snakes 
c. *a frightened man by snakes 
d. *an angry man at John 
(35) a. a man eager to please 
b. a man fearful of snakes 
c. a man frightened by snakes 
d. a man angry at John 
292 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
The above examples might suggest that what distinguishes easy from these other 
adjectives is that the VP/NP complement of easy is obligatory, while the PP comple- 
ments of the above adjectives are optional. While there are not many adjectives against 
which to test this hypothesis, the one clear case of an adjective that takes an obligatory 
complement counts against the idea: 
(36) a. a man fond of snakes 
b. *a fond man 
c. *a fond man of snakes. 
The analysis we propose localizes in lexical entries the ability of a subcat to 
be transferred from adjunct to head. Just as subcats can be marked for the oblig- 
atory/optional distinction in a class definition or in a lexical entry, so can they be 
marked for a distinction we term transferable. While as a default subcats will be non- 
transferable, those subcats that are identified by a class or lexical entry as transferable 
will be subject to the following informally stated principle. 
Transferable Subcat Principle. When a transferable subcat on a daugh- 
ter in a local subtree is not associated with some sister in that subtree, 
the subcat becomes part of the corresponding subcat list of the head 
daughter in that subtree. 
In the constructions studied here, this principle applies in cases where the lexical 
entry or phrase with a transferable subcat serves as an adjunct (easy) or a specifier (too), 
so that the word or phrase's subcat list is not used directly. The intent of the principle in 
such cases is to make the transferable subcat a part of the head, so the subcategorization 
principle will ensure that the information is propagated to the mother node. This is 
intended as a modification of the subcategorization principle---note that it has the 
effect of licensing a kind of "discontinuous constituent. "~7 
Having introduced this additional property of subcats, that they can be specified 
as transferable, we note that the default value for this property must be negative, since 
in general subcats from adjuncts and specifiers do not pass to heads, as seen in (33) 
and (34) above. This default value will be overridden for the VP/NP and the For- 
PP subcats in the SLASH-COMP class, to reflect the grammaticality of both examples 
in (37). 
(37) a. That was a melodious sonata to listen to. 
b. John is an easy man to please. 
Members of the SLASH-COMP class, including the relevant lexical entry for easy, 
will inherit this nondefault transferable property for both the XComp and the For-PP, 
so when easy combines as an adjunct with the head noun man, these two subcats 
will become part of the subcategorization of man, by the principle above, and will 
then become part of the subcategorization of the node easy man, accounting for the 
grammaticality of (31) above. 
Example (32a) will be ruled out because of an independent constraint that re- 
stricts pre-head adjuncts to those that are (Head-Final +). Example (32b) is excluded 
because easy has an obligatory VP/NP complement, which must be included as an 
17 The ability to transfer a subcategorization requirement from a modifier to a mother (or to a head) is 
perhaps a bit similar in effect to FUNCTION COMPOSITION in categorial grammar (Bach 1988). But in 
HPSG the possibility may be lexically constrained. 
293 
Computational Linguistics Volume 18, Number 3 
obligatory complement of the phrase easy man, due to the convention adopted above 
about merging of subcat information between a head and its sister. Finally, (33a) is 
excluded because the easy that requires an unslashed VP complement will not pass 
on its XComp subcat to the noun it modifies, since that XComp is, like most subcats, 
nontransferrable. So easy man to please Bill will be excluded for the same reason that 
eager man to please is excluded: nothing licenses the postnominal infinitival VP. 
Example (33b) is probably best excluded on semantic grounds, since the subject 
of easy to please Bill is an expletive pronoun, the wrong sort to unify with the head 
noun being modified. On the assumption that a noun must serve semantically as the 
subject of adjectival adjuncts, those adjuncts must specify some thematic role for the 
noun to play. Thus any adjective that requires an expletive subject should give rise to 
a semantically ill-formed expression when it appears as an adjunct to a noun. What 
prevents the IT-EASY easy from serving as an adjunct to problem is the fact that this 
easy requires an expletive it subject. 
Given this transferable complement mechanism, we may straightforwardly com- 
plete the analysis of the earlier too/enough examples: the lexical entries for these two 
adverbs simply specify that their gappy infinitival complement is transferable. TM 
6. Conclusions and Future Directions 
The study of inheritance and, more generally, the study of structured lexical repre- 
sentations is an exciting and promising field. We would like to use this section to 
summarize how we view this work and to suggest directions in which we feel it 
should move. 
6.1 Conclusions 
We have presented a treatment of complementation that uses nonmonotonic lexical 
inheritance. The lexical specifications are quite compact and therefore both readily 
extendible and easily modified. We pointed out cases where nonmonotonic, default 
specification seems most natural, and the entire treatment turns on the possibility of 
there being genuine multiple inheritance of a "complements" attribute. 
We adopt a skeptical approach to inheritance conflict. If there are inheritance con- 
flicts in the system presented here, nothing is inherited. Mechanisms that warn users 
about such conflicts are useful, but we are wary of attempts to decide conflicts "in- 
telligently." They seem likely to us to lead to cases where minor changes may have 
remote consequences, which would detract from maintainability. 
We do not feel that we have overstated the case for structured lexicons by choosing 
a particularly messy or poorly understood area. To insist on this point somewhat, let 
us note that we omitted significant aspects of the grammar of the "raising nouns," e.g., 
their complements, specifiers, and adjuncts. 19 Grammar abounds in poorly understood 
areas, including comparatives, superlatives, adverbials, internal NP syntax, and the 
"specialized grammars" found in dates, places, and technical vocabulary. All of these 
areas can benefit from the application of a tool for complex lexical description. 
18 This leaves much to be said about the lexical properties of too and enough, but more detailed analysis at 
this point would take us too far afield; it is clear enough that, whatever their other properties, these 
two adverbs share complementation properties with the adjectives and nouns studied here. 
19 For example, for "pleasure" nouns, some adjectives are okay, but not all (a real/*competent pleasure to 
work with; relative clauses are impossible: Sally is a pleasure *\[that is real\] to work with; and some nominal 
complements are fine: Sally was a pleasure of the rarest kind to work with. 
294 
Dan Flickinger and John Nerbonne Inheritance and Cornplernentation 
6.2 Hypotheses or Tools? 
The conclusions above may be read as a plea for the employment of an important tool 
in computational linguistics, and, indeed, we see the primary significance of the use of 
structured lexicons not in new expressive power which they bring to natural language 
processing or description (there is perhaps none), but rather in the increased ease and 
reliability with which they allow old hypotheses to be formulated and put to use. 
Brachman (1983, p. 35) summarizes the dominant view of inheritance in knowl- 
edge representation: 
Even though much has been made in the past of the significance of in- 
heritance in semantic nets, no one has been able to show that it makes 
any difference in the expressive power of the system that advertises 
it... It is strictly implementational. 
Given this authority on the technical side, it may be surprising to hear more 
application-oriented users of inheritance mechanisms hedging at all on whether there 
is any scientific significance to the proposal here. But there is at least a potential 
candidate: lexical rules may distinguish inherited from specified information. 
In expressing the relationships between members of two sets of lexical entries, 
we make crucial use of the distinction between idiosyncratically specified information 
(which appears in a sparse Inonredundant\] lexical entry) and inherited information. We 
have adopted here the restrictive hypothesis, proposed in Flickinger (1987), that lexical 
rules hold for minimally specified lexical entries, without having access to inherited, 
predictable information. Adopting this hypothesis imposes a constraint on the form 
and function of lexical rules that is strong, perhaps too strong, but one that allows a 
simpler formulation of rules by keeping to a minimum the amount of information to be 
managed. Only two kinds of information are relevant for a lexical rule: the word classes 
that each of the two related entries belong to, and any idiosyncratic properties specified 
by either lexical entry. We note that if lexical rules were insensitive to the distinction 
between idiosyncratic and predictable properties of lexical entries, the statement of 
even a simple rule like LR-PAST, given earlier, would be much more difficult. If the 
lexical rule for past tense verbs had to cope with fully specified entries that blurred 
this distinction, it would be difficult to express in the rule just which properties of the 
one entry had to match in the other, related entry. 
For example, the verb like idiosyncratically requires a verbal complement that is 
either infinitival or gerundive, while the verb enjoy does not allow the infinitival form, 
allowing only the gerundive form for its complement. Since all of the inflected forms of 
like allow the same choice of two permissible forms for the verbal complement, while 
all of the inflected forms of enjoy insist on the gerundive complement, the lexical rules 
like LR-PAST or the similar one for present third singular forms must preserve these 
idiosyncracies. Yet a fully specified entry for the base form enjoy stipulates not just 
the form of the complement, which would have to be identical in the present third 
singular entry enjoys; the fully specified base entry for enjoy also specifies that its 
subject be unmarked for number, an indifference that crucially must not be shared by 
the entry for enjoys. Short of tagging each attribute value in a fully specified entry as 
local or inherited, it is not clear how the lexical rule for present third singular forms 
could be constrained to ensure identity of the verbal complement's VFORM value 
while ignoring differences in the subject's AGREEMENT value for these two entries 
for enjoy. In sharp contrast, this difference in idiosyncratic vs. inherited information 
can be exploited by lexical rules without stipulation when they are constrained to 
apply only to minimally specified entries. 
295 
Computational Linguistics Volume 18, Number 3 
It may not be superfluous to add that, even if the argument above about distin- 
guishing inherited and specified information is ultimately fallacious, so that the use 
of inheritance were seen purely as a tool, and not at all as a scientific hypothesis, it 
may nonetheless prove to be of great significance, just as many tools have advanced 
areas of science that nothing to do with their development. The development of lenses 
revolutionized astronomy, even though glass grinding embodied no astronomical hy- 
potheses. 
6.3 Emergent Issues in Structured Lexicons 
Perhaps more interesting are the many directions in which this research may be de- 
veloped. We suggest some of these in the questions below. 
What are lexical classes and lexical entries? The careful reader noted in Section 5 
above that our lexical specifications are translated into feature structures. Theoretically, 
we could dispense with the translation for nearly all of the information involved, and 
have the lexicon describe feature structures directly. But this does not correspond to 
our implementation, nor are we clear on how, e.g., information on lexical rules and 
their application ought to be rendered in features. Perhaps lexical entries must be 
structured so that one component of a lexical entry is a feature structure, while others 
are not. 2° 
Can inheritance be exploited in the specification of inflectional variation? This 
appears to be a promising area of application, since in general, one can view inflected 
elements as further specifications of abstract lexemes (cf. Evans and Gazdar 1989 for 
an intriguing proposal). 
Can derivational lexical rules be treated more satisfactorily? For example, it is 
clear that at least some lexical rules relate not merely a pair of word classes, but 
rather entire lexical substructures (involving several classes) to one another. Can the 
techniques of inheritance be applied here, so that exceptional elements may be easily 
accommodated? 
Acknowledgments 
We are indebted to Mark Gawron, Masayo 
Iida, Bill Ladusaw, Joachim Laubsch, Carl 
Pollard, and Tom Wasow for frequent 
conversations about this analysis. We are 
also grateful to Anthony Kroch, the 
participants at the Tilburg Workshop on 
Inheritance in Natural Language Processing, 
and three referees for further comments. 
This work was partially supported by a 
research grant, ITW 9002 0, from the 
German Bundesministerium ftir Forschung 
und Technologie to the DFKI DISCO project. 
References 
Bach, Emmon. (1988). "Categorial grammars 
as theories of language." In Categorial 
Grammars and Natural Language Structures, 
edited by Richard T. Oehrle, Emmon 
Bach, and Deirdre Wheeler. Reidel. 
Baltin, Mark. (1987). "Degree complements." 
In Syntax and Semantics 20: Discontinuous 
Constituency, edited by Geoffrey J. Huck 
and Almerindo E. Ojeda. Academic Press. 
Bayer, Samuel. (1990). "Tough movement as 
function composition." In Proceedings of 
the 9th West Coast Conference on Formal 
Linguistics, edited by Aaron Halpern. 
CSLI, Stanford, 29--42. 
Borsley, Robert. (1983). "A Welsh agreement 
process and the status of VP and S." In 
Order, Concord, and Constituency, edited by 
Gerald Gazdar, Ewan Klein, and Geoffrey 
Pullum. Foris Publications. 
Borsley, Robert. (1987). "Subjects and 
complements in HPSG." CSLI Report 
No. CSLI-87-107, Center for the Study of 
Language and Information, Stanford 
University, Stanford, CA. 
Brachman, Ronald J., and Schmolze, James 
G. (1985). "An overview of the KL-ONE 
knowledge representation system." 
Cognitive Science, 9, 171-216. 
20 Krieger and Nerbonne (1991) attempt to characterize all lexical information (including inflectional and 
derivationat relationships) in typed feature structures--with no further information. 
296 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
Brachman, Ronald J. (1983). "What IS-A is 
and isn't: An analysis of taxonomic links 
in semantic networks." IEEE Computer, 
30-36. 
Brame, Michael. (1979). Essays Toward 
Realistic Syntax. Noit Amrofer. 
Bresnan, Joan. (1971). "Sentence stress and 
syntactic transformations." Language, 47, 
257-281. 
Bresnan, Joan (ed.) (1982). The Mental 
Representation of Grammatical Relations. MIT 
Press. 
Carpenter, Bob, Pollard, Carl J., and Franz, 
Alex. (1991). "The specification and 
implementation of constraint-based 
unification grammar." Unpublished 
report, CMU Laboratory for 
Computational Linguistics. 
Chomsky, Noam. (1965). Aspects of the Theory 
of Syntax. MIT Press. 
Chomsky, Noam. (1973). "Conditions on 
transformations." In A Festschrift for Morris 
Halle, edited by Stephen Anderson and 
Paul Kiparsky. Holt, Rhinehart and 
Winston. 
Chomsky, Noam. (1977). "On wh 
movement." In Formal Syntax, edited by 
Peter Culicover, Tom Wasow, and Adrian 
Akrnajian. Academic Press. 
Culicover, Peter, and Wilkens, Wendy. 
(1984). Locality in Linguistic Theory. 
Academic Press. 
Devlin, Keith. (1991). Logic and Information. 
Oxford University Press. 
Emele, Martin, and Zajac, Remi. (1990). 
"Typed unification grammars." In 
Proceedings, 13th International Conference on 
Computational Linguistics, Helsinki. 
293-298. 
Evans, Roger and Gazdar, Gerald. (1989a). 
"Inference in DATR." In Proceedings, 4th 
Conference of the European Chapter of the 
Association for Computational Linguistics. 
66-71. 
Evans, Roger and Gazdar, Gerald. (1989b). 
"The Semantics of DATR." In Proceedings, 
7th Conference of the Society for the Study of 
Artificial Intelligence and Simulation of 
Behaviour. Pittman/Morgan Kaufmann, 
London, 79-87. 
Fillmore, Charles. (1985). "Pragmatically 
controlled zero anaphora." In Proceedings, 
7th Meeting of the Berkeley Linguistic Society. 
Flickinger, Daniel. (1987). Lexical rules in 
the hierarchical lexicon. Doctoral 
dissertation, Stanford University, 
Stanford, CA. 
Flickinger, Daniel, Pollard, Carl, and 
Wasow, Tom. (1985). "Structure-sharing in 
lexical representation." In Proceedings, 23rd 
Annual Meeting of the Association for 
Computational Linguistics, 262-267. 
Fodor, Janet D. (1978). "Parsing strategies 
and constraints on transformations." 
Linguistic Inquiry, 9: 427-473. 
Franz, Alex. (1990). '% parser for HPSG." 
Technical Report LCL-90-3, CMU 
Laboratory for Computational Linguistics. 
Gazdar, Gerald; Klein, Ewan; Pullum, 
Geoffrey; and Sag, Ivan. (1985). 
Generalized Phrase Structure Grammar. 
Harvard University Press. Cambridge, 
MA. 
Goldstein, Ira and Roberts, B. (1977). "The 
FRL Manual." MIT-AI Memo 409, MIT. 
Cambridge, MA. 
Gunji, Takao. (1987). Japanese Phrase Structure 
Grammar. Reidel. 
Hukari, Tom, and Levine, Robert. (1991). 
"On the disunity of unbounded 
dependency constructions." Natural 
Language and Linguistic Theory, 9: 97-144. 
Jackendoff, Ray. (1972). Semantic 
Interpretation in Generative Grammar. MIT 
Press. 
Jackendoff, Ray. (1975). "Tough and the trace 
theory of movement rules." Linguistic 
Inquiry, 6: 437-446. 
Jacobson, Pauline. (1982). "Evidence for 
gaps." In The Nature of Syntactic 
Representation, edited by Pauline Jacobson 
and Geoffrey K. Pullum. Reidel, 187-228. 
Jacobson, Pauline. (1984). "Connectivity in 
Phrase Structure Grammar." Natural 
Language and Linguistic Theory, 1: 535-581. 
Jacobson, Pauline. (1987). "Phrase structure, 
grammatical relations, and discontinuous 
constituents." In Syntax and Semantics 20: 
Discontinuous Constituency, edited by 
Geoffrey J. Huck and Almerindo 
E. Ojeda. Academic Press. 
Jacobson, Pauline. (1990). "Raising as 
function composition." Linguistics and 
Philosophy, 13(4): 423--476. 
Johnson, Mark. (1988). Attribute Value Logic 
and the Theory of Grammar. CLSI. 
Jones, Charles. (1990). "Decapitation (of 
some so-called 'Null-Operator 
Constructions')." In Proceedings of the 9th 
West Coast Conference on Formal Linguistics, 
edited by Aaron Halpern. CSLI, Stanford, 
317-30. 
Kaplan, Ronald, and Bresnan, Joan. (1982). 
"Lexical-functional grammar: A formal 
system for grammatical representation." 
In The Mental Representation of Grammatical 
Relations, edited by Joan Bresnan. MIT 
Press. 
Krieger, Hans-Ulrich, and Nerbonne, John. 
(1991). "Feature-based inheritance 
networks for computational lexicons." 
DFKl Research Report RR-91-31, 
297 
Computational Linguistics Volume 18, Number 3 
Deutsches Forschungszentrum f~ir 
Kiinstliche Intelligenz, SaarbrUcken, 
Germany. (Also in: Ted Briscoe, Anne 
Copestake, and Valeria de Paiva, eds., 
Proceedings of the ACQUILEX Workshop on 
Default Inheritance in the Lexicon. Technical 
Report No. 238, University of Cambridge 
Computer Laboratory, Cambridge, U.K.) 
Lasnik, Howard, and Fiengo, Robert. (1974). 
"Complement object deletion." Linguistic 
Inquiry, 5: 535-571. 
Maling, Joan, and Zaenen, Annie. (1982). "A 
phrase structure account of Scandinavian 
extraction phenomena." In The Nature of 
Syntactic Representation, edited by Pauline 
Jacobson and Geoffrey K. Pullum. Reidel. 
Miller, George, and Chomsky, Noam. (1963). 
"Finitary models of language users." In 
Handbook of Mathematical Psychology Vol. II, 
edited by R. D. Luce, R. Bush, and 
E. Galanter. Wiley. 
Montague, Richard. (1973). "The proper 
treatment of quantification in ordinary 
English." In Formal Philosophy, edited by 
Richmond Thomason. Yale University 
Press, 247-270. 
Nanni, Deborah L. (1980). "On the surface 
syntax of constructions with easy-type 
adjectives." Language, 56: 568-81. 
Nerbonne, John, and Proudian, Derek. 
(1987). "The HP-NL system." Technical 
report, Hewlett-Packard Labs, Palo Alto, 
CA. 
Nerbonne, John. (1992). "A feature-based 
syntax/semantics interface." In 
Proceedings, Second ConfErence on the 
Mathematics of Language} edited by Alexis 
Manaster-Ramer and~Wlodek Zadrozny. 
Annals of Mathematics and Artificial 
Intelligence. 
Pollard, Carl. (1984). Head grammars, 
generalized phrase structure grammars, and 
natural language. Doctoral dissertation, 
Stanford University, Stanford, CA. 
Pollard, Carl. (1985). "Phrase structure 
grammar without metarules." In 
Proceedings, 4th Annual Meeting of the West 
Coast Conference on Formal Linguistics. 
CSLI, Stanford, 246-261. 
Pollard, Carl. (1988). "Categorial grammar 
and phrase structure grammar: An 
excursion on the syntax-semantics 
frontier." In Categorial Grammars and 
Natural Language Structures, edited by 
Richard T. Oehrle, Emmon Bach, and 
Deidre Wheeler. Reidel. 
Pollard, Carl. (1989). "The syntax-semantics 
interface in a unification-based phrase 
structure grammar." In Views of the 
Syntax-Semantics Interface: Proceedings of the 
Workshop " GPSG and Semantics," edited by 
Stephan Busemann, Christa Hauenschild, 
and Carla Umbach. Technische 
Universit~it Berlin, 22-24. Feb 1989, 
167-184. KIT FAST, Technische 
Universit~it Berlin. 
Pollard, Carl, and Sag, Ivan. (1987). An 
Information-Based Theory of Syntax and 
Semantics, Vol. I. CSLI, Stanford. 
Pollard, Carl, and Sag, Ivan. (1988). "An 
information-based theory of agreement." 
In Proceedings from the Parasession on 
Agreement, edited by Diana Brentari, Gary 
Larson, and Lynn McCleod. Chicago 
Linguistics Society, Chicago, 236-257. 
Pollard, Carl, and Sag, Ivan. (1991). "An 
information-based theory of syntax and 
semantics." Vol. II. Unpublished 
manuscript in preparation. 
Postal, Paul. (1971). Cross-over Phenomena. 
Holt, Rinehart, and Winston. 
Proudian, Derek, and Pollard, Carl. (1985). 
"Parsing head-driven phrase structure 
grammar." In Proceedings, 25th Annual 
Meeting of the Association for Computational 
Linguistics, 167-71. 
Purdy, William C. (1988). "A lexical 
extension of Montague semantics," 
Report CIS-88-1, School of Computer and 
Information Science, Syracuse University. 
Roberts, Diana. (1991). "Linking rules in an 
HPSG lexicon." Master's thesis, CorneU 
University. 
Rosenbaum, Peter S. (1967). The Grammar of 
English Complement Constructions. MIT 
Press. 
Ross, John R. (1967). Constraints on Variables 
in Syntax. Doctoral dissertation, MIT, 
Cambridge, MA. 
Sag, Ivan A. (1982). "A semantic theory of 
'NP movement' dependencies." In The 
Nature of Syntactic Representation, edited by 
Pauline Jacobson and Geoffrey K. Pullum. 
Reidel. 
Schachter, Paul. (1981). "Lovely to look at." 
Linguistic Analysis, 8: 431-48. 
Shieber, Stuart. (1986). An Introduction to 
Unification-Based Approaches to Grammar. 
CSLI, Stanford. 
298 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
Appendix A: Lexical Fundamentals 
In the analysis of adjectives governing VP complements that we provide here, we adopt 
the approach to lexical representation developed in Flickinger (1987). We provide here 
a brief sketch of the framework, limiting our discussion to those aspects that are 
relevant to the analysis in the main body of the paper. 
Lexical Framework 
Much of the information in a fully specified entry within the lexicon is not unique to 
that particular entry. Viewing this information as a set of discrete properties making 
up the lexical entry, related lexical items will have in common some of the properties 
of that entry. Lexical items can be grouped into classes defined by those properties 
that are common to all members of the class. By giving a precise characterization 
of these word classes, one can eliminate a good deal of the redundancy found in a 
lexicon that consisted of fully specified entries. Put differently, one can make use of 
these word classes to capture generalizations about the elements of the lexicon and to 
make predictions about the behavior and distribution of a lexical item on the basis of 
the classes it belongs to. 
To avoid redundancy entirely, each property relevant to representing the elements 
of a lexicon should only be mentioned once in some single class, with all elements of 
the lexicon that have this property being members of that class. If so, a given lexical 
item may have to belong to many classes in order to obtain all of its properties. These 
word classes form a hierarchy over which rules are defined that govern the inheritance 
of information from class to class, and ultimately to individual lexical entries. One of 
the nicest aspects of this idea is that it is readily visualized (cf. Figure 7). 
In order to present the class definitions that formally express the properties of the 
adjectives governing VP complements as collections of attributes with assigned values, 
we shall make use of the hierarchy of word classes above, where classes inherit their 
properties from other more general classes, or indeed from several such classes. This 
hierarchy of classes is intended as the repository of both the syntactic and semantic 
properties that comprise the fully fleshed-out lexical entries of a language; we will of 
course only present a very few of the classes that would populate a complete hierarchy, 
sufficient we hope to prove the promise of such a representation. Part of this hierarchy 
is devoted to specifying the number and types of complements that predicates allow 
or require, and among this group of word classes is one defining the properties shared 
by lexical entries that take a subject and a complement that is semantically controlled 
(possibly by the subject); we label this class simply CONTROL. Above, we illustrate 
the barest outlines of the top of the word class hierarchy for English, to suggest where 
the CONTROL class fits in. We provide the content of CONTROL later. 
It is important to note that word classes do not merely form a simple hierarchy, but 
rather that they form a set of interconnected hierarchies; as Chomsky (1965) argued in 
his discussion of lexical representation in Aspects (pp. 79-83), the need to cross-classify 
a given lexical item according to several distinct properties renders impossible the use 
of a simple branching hierarchy to represent the lexicon. 
We take as adequate motivation for the existence of a word class the demonstration 
that some particular syntactic or morphological property (or cluster of properties) is 
shared by a number of lexical items. The forcing function in the identification of word 
classes is our assumption that the best representation for lexical information is one 
in which each new piece of information, each distinct property exhibited by one or 
more elements of the lexicon, is introduced exactly once in the lexicon. A property 
299 
Computational Linguistics Volume 18, Number 3 
WORD 
COMPLEMENTATION / 
INCOMPLETE 
PART-OF-SPEECH 
COMPLETE ...... 
CONTROL TRANSITIVE 
Figure 7 
A sample of lexical classes near the top of the lexical hierarchy. 
shared by more than one lexical item should be introduced in a word class common 
to those items, or those lexical items should be related by lexical rule. In the simplest 
case where a word class or lexical entry belongs to just one superclass, the inheritance 
rule for how values of an attribute are assigned in the word class hierarchy is quite 
straightforward. The two alternatives to be considered depend on whether a given 
attribute permits only one value, or multiple values. 
Inheritance of Values 
The value assigned to a particular word class (or member) W for a 
given attribute is determined as follows: 
a. For a single-valued attribute, the assigned value is either 
introduced directly in W, or is the one introduced in the 
most specific class to which W belongs. If there is no value 
introduced anywhere in the linked classes between W and 
the root WORD-CLASS, inclusively, no value is assigned to 
W for that attribute. 
b. For a multiple-valued attribute, the assigned values are the 
members of the set consisting of all distinct values 
300 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
introduced for that attribute in W and in any of the classes 
linking W with the root WORD-CLASS, inclusively. 2~ 
In cases where a class or member belongs to more than one superclass, the picture 
might be more complicated, since each of two immediate superclasses might specify a 
different value for the same single-valued attribute. One way to address the potential 
conflict would be to define another rule of inheritance to take account of multiple 
parents, a rule which for each attribute assigns priority to some one of the parent 
classes (see the paper on the ELU lexicon in this collection). We instead assume that 
the hierarchy is specified in such a way that each single-valued attribute of some given 
class or member was assigned a value by at most one of the immediate superclasses 
(or its parents), so conflicting values do not occur. 
In adopting this discipline for inheritance of lexical information, we follow Flick- 
inger (1987) rather than accepting the more rigorously defined but more restrictive 
rules of inheritance defined for DATR by Evans and Gazdar (1989a; 1989b). As will be 
seen in the discussion to follow, we believe that the ability to inherit a lexical property 
from more than one potential source can be important in capturing relevant linguis- 
tic generalizations. There would be no point to our use of multiple-valued attributes 
if we did not allow genuine multiple inheritance. We employ the multiple-valued 
attribute "complements" in order to collect subcategorization information about sev- 
eral subcategorized-for elements simultaneously. In its most natural form, the single 
property "complements" inherits from several ancestors simultaneously. 22 
We also follow Flickinger, this time accompanied by Evans and Gazdar, in as- 
suming that lexical attributes may be assigned default values as part of a word class 
definition, with those defaults possibly being overridden in the definition of a subclass 
or lexical entry inheriting from that word class. In this we part ways with Pollard and 
Sag (1987)'s strong assumption of monotonicity in the inheritance of lexical properties, 
again for reasons that we identify in the discussion to follow. 23 
While word classes and the associated mechanism of nonmonotonic inheritance 
provide powerful tools for representing one kind of shared information, that which 
links a category to its subcategories, a distinct formal device is required to link two 
morphologically related classes of different categories. We employ the familiar notion 
of a lexical rule to represent this second kind of systematic (but not exceptionless) 
relationship. Given a word belonging to the first set, a lexical rule predicts the exis- 
tence of a corresponding word belonging to the second set, with the differences and 
similarities between the two words captured both in the formulation of the rule, and 
in the definitions of the classes each word is a member of. 
We illustrate our notation for lexical rules with an example of a relatively simple 
one, the inflectional rule relating the base forms of verbs with their past tense forms. 
We can express the relationship between, say, walk and walked as given in the LR-PAST 
rule below, leaving out specifics of phonology and only hinting at semantics. 
21 We employ multiple-valued attributes only within the lexicon in order to gather subcategorization 
specifications. A translation step converts these to sets (or--with more information--lists) for use in the 
feature system. 
22 The DATR position carefully disallows multiple inheritance of a single property from two or more 
classes, even while allowing inheritance from various classes into different properties (in a single word 
class). We find genuine multiple inheritance seems useful, even if it may be dispensible; cf. the 
treatment of the lexical attribute "complements" below. 
23 But see Pollard and Sag 1987, p. 194 note. 
301 
Computational Linguistics Volume 18, Number 3 
Rule 
LR-PAST lexical rule 
LR-PAST 
LE2-Classes - PAST = LE1-Classes - BASE 
LE2-Spelling = (AFFIX-ED LE1-Spelling) 
LE2-Phonology .... 
LE2-Semantics = (PAST LE1-Semantics) 
This lexical rule, like any other, expresses a relation holding between two sets of 
lexical entries, the first set represented by a canonical lexical entry LE1, and the second 
set by LE2. The rule's applicability is governed by the relevant classes that LE1 and 
LE2 each belong to, with these classes named in the statement within the rule that 
relates the one entry's list of parent classes with the other entry's class list. Having 
specified the range of applicability, each rule then states the particular dependencies 
holding between properties of LE1 and corresponding properties of LE2. 24 
We might view inheritance within a hierarchy of word classes as a tool that elimi- 
nates redundancy along one dimension within the lexicon, while lexical rules provide 
the same service along another dimension. A given lexical item, by virtue of being a 
member of one or more word classes, shares inherited properties with other lexical 
items that belong to those same classes, but does not necessarily share a common 
morphological or semantic base (or indeed any idiosyncratic properties) with any of 
those other items. That same lexical item, by participating in one or more lexical rules, 
has properties in common with a second set of lexical items, where the shared prop- 
erties crucially include some or all of the idiosyncratic information that distinguishes 
the lexical item from others in its class. The members of this second set, related by 
lexical rules, all do share a single common semantic and morphological base (except 
for suppletions). 
Of course, if a lexical rule relates two entries that both belong to a given word 
class (as happens with the verbal inflection rules), those two entries will share some 
inherited properties as well as the idiosyncracies. However, the lexical rule only es- 
tablishes joint membership in that given class and the relationship of the idiosyncratic 
information in the two entries; all other properties shared by the two are established 
by inheritance within the word class hierarchy. 
Both of these formal devices, inheritance and lexical rules, serve to express that 
which is common among (often overlapping) sets of fully specified lexical entries, 
including properties that are morphological, syntactic, and semantic. In their capacity 
as redundancy mechanisms, the two devices permit a parsimonious representation of 
the existing lexicon. 
Lexical Content 
For convenience of exposition, we view the syntactic properties of lexical items as 
being of two kinds: one a set of features, separated into those with atomic values and 
those with category values, and the other a set of subcategorization specifications, di- 
vided into complements (obligatory and optional), and adjuncts. Even though HPSG 
represents both types of information (features and subcategorization specifications) 
24 We include this rule primarily in order to introduce the notation we shall later employ. We would, 
however, be quite sympathetic to an alternative treatment of the relation between PAST tense forms 
and untensed lexemes that employed lexical inheritance rather than inflectional rules to account for the 
relationship. We deny that this sort of treatment can be extended straightforwardly from inflectional to 
derivational rules, however. 
302 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
uniformly as attribute-value pairs (as noted above), we shall represent them lexicaIIy 
as distinct. We have two motivations for this: first, in representing the information 
differently from its normal form in HPSG we demonstrate the independence of the 
lexical ideas presented here. We employ HPSG in the grammatical analysis presented 
here because it is a useful grammatical framework, and because it makes strenuous 
lexical demands; but the lexicon framework does not presuppose HPSG (for exam- 
ple, PATR-II systems can make use of structured lexicons, and nearly do, in the form 
of templates. Compare Shieber 1986, pp. 54-55.). Second, a uniform feature notation 
needed for subcategorization and nonsubcategorization information threatens to ob- 
scure points addressed below, so the two kinds of information are separated in the 
representations of lexical entries used here. 
The atomic-valued features that we employ in specifying lexical entries (and in 
specifying categories) are drawn from a (small) finite set where each feature has a 
limited set of possible atomic values, 25 e.g., the binary feature INVERTED, indicating 
whether or not a verb can appear as the head of an inverted sentence. The feature 
VFORM, on the other hand, draws its values from a set containing among others 
BASE, PAST, and PAST-PARTICIPLE, to represent the morphological form of a verb. 
Category-valued features take as their value a feature structure, a specification for 
some syntactic category. Since any nonempty set of feature value pairs is (by definition) 
enough to specify a category, any such set constitutes a possible value for one of these 
category-valued features. 
Each complement or adjunct entry, referred to here as a subcat, consists of a cat- 
egory specification and its semantic properties. Since reference to subcategorization 
properties of subcats is excluded in specifying complements or adjuncts within a lex- 
ical entry (cf. Pollard and Sag 1987:143-4 for a similar--but not identical--"locality" 
restriction), we make use of a feature COMPLETE, quite similar to the SUBJ fea- 
ture proposed by Borsley (1983), and employed in Gazdar et al. (1985:61f) to distin- 
guish incomplete from complete categories. Incomplete constituents lack one or more 
of their obligatory complements, including at least their final complement (usually 
the subject), and are marked \[COMPLETE-\], while complete categories are marked 
\[COMPLETE+\] to represent the property that no obligatory arguments are missing. 
(Complete categories correspond roughly to maximal projections in an X-bar frame- 
work.) To distinguish lexical categories from phrasal ones, we use the binary feature 
LEXICAL. With these two features COMPLETE and LEXICAL, we can follow Pollard 
(1984) in dispensing with the widely used (and abused) X-bar machinery, while main- 
taining the full range of necessary distinctions among lexical and phrasal categories. 
The content of the word-class CONTROL, promised above, is presented here. 
(38) 
CONTROL 
Superclasses 
Complements 
XComp-Features 
XComp-Subj-Semantics 
XComp-Oblig 
XComp-Semantics 
XComp-Role 
Incomplete 
XComp 
(Category Verb)(Complete -) 
(Lexical -) 
Subject-Semantics 
Yes 
XComp-Semantics 
State-of-Affairs 
25 For linguistic defense of many of the actual features used here, see (Gazdar et al. 1985). 
303 
Computational Linguistics Volume 18, Number 3 
As the hierarchy of word classes sketched above indicates, this class inherits from 
the INCOMPLETE class (which specifies an obligatory subject complement), and in- 
troduces a second obligatory complement that is a verb phrase (not complete, which 
would be its maximal projection, a sentence; and not lexical, which would be just 
a verb without any complements). It will play the role State-of-Affairs (abbreviated 
'soa') in relations denoted by words inheriting from CONTROL, and it will be semanti- 
cally interpreted by the variable XComp-Semantics. The specification of the semantics 
will occasionally be omitted below, since the convention should be clear. It is this 
CONTROL class that will serve as the superclass from which both of the adjectival 
VP complement classes (cf. IT-EASY and SLASH-EASY below) inherit. 
Before concluding the sketch of the lexicon, we turn to the lexical representation 
of semantics, which likewise plays a role in the final analysis. 
Lexical Semantics 
The use of hierarchies of classes of information, advocated here as a means of rep- 
resentation for grammatical information, is also common in the representation of se- 
mantic hyponymy relations, e.g., the relation between boy and child (cf. Brachman and 
Schmolze 1985). Thus boy is a hyponym (subconcept) of both child and male, which are 
in turn hyponyms of more abstract words and concepts. This may be modeled in the 
same multihierarchical fashion we employ for grammatical information. Such seman- 
tic hierarchies may be of utility in constructing more efficient NL inference engines 
(Purdy 1988). 
We exploit a very different use of semantic inheritance in the present treatment, 
however, beginning with the observation familiar from categorial grammar (Bach 1988) 
that semantics and subcategorization are interdependent: subcategorizers denote re- 
lations among the denotations of their complements. Montague (1973) effectively ex- 
ploited this by interpreting multiplace verbs and verb phrases as functions into the 
denotations of lesser-place verbs. We exploit the interdependence by allowing some 
semantic inheritance to follow syntactic subcategorization lines. To be more precise, 
we allow subcategorizers to specify not only the syntax of their complements, but also 
the semantic role the complement is assigned in the relation denoted by the subcate- 
gorizer. 
INCOMPLETE 
Superclasses Complementation 
Complements Subject 
Subject-Features (Complete+)(Category Noun) 
Subject-Role Source 
TRANSITWE 
Superclasses Incomplete 
Complements Object 
Object-Features (Complete+)(Case Accusative)(Category Noun) 
Object-Role Theme 
Thus INCOMPLETE assigns its subject the role source; TRANSITIVE inherits this role 
assignment and extends it by assigning theme to objects. 
Roles may be understood by their function in atomic formulas: in standard pred- 
icate logic the binding of arguments to argument positions is mediated by the order 
304 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
in which arguments appear. Rxy ~ Ryx. The use of explicit roles in semantic rela- 
tions accomplishes this task and obviates the order of arguments: R(source:x, theme:y) 
= R(theme:y, source:x). Role-coded formulas are more easily readable when there 
are many roles, and the use of roles seems essential in semantic theories of topics 
such as variable-place relations or variably-binding arguments such as the possessive. 
There is furthermore a substantial body of work on the so-called "linking" of seman- 
tic roles to syntactic information, including especially Roberts (1991) who applies this 
theory to HPSG. (Even though we use role- or keyword-coded arguments in lexical 
specifications, we will occasionally revert to order-coded representations for the pur- 
poses of illustration. They are more concise. Compare Nerbonne (1992) for discussion 
of the semantic status of roles.) 
What is important about roles for the present application is that we may exploit 
the inheritance mechanism to derive (specifications for) semantics for multiplace sub- 
categorizers. Instances of the TRANSITIVE class are assigned the following semantics, 
using the multi-valued inheritance scheme discussed above. 
pred: 
source: 
theme: 
Subject-Semantics \] Object-Semantics 
Lexical specifications for the arguments to the roles have not been shown, but the 
general scheme should be clean The predicate must of course be assigned by each 
individual lexical entry. "Subject-Semantics" is used because it is useful to be able to 
refer to the semantics of a given complement (the controller) in cases of grammatical 
control, as the semantics specifications for the lexical class CONTROL demonstrate. 
Exceptions to Lexical Rules 
Since few if any lexical rules prove to be completely exceptionless, we assume that 
individual lexical entries can and do stipulate among their idiosyncratic properties 
exceptional behavior with respect to particular lexical rules. One such exceptional 
property is that a given entry belongs to a class to which a lexical rule applies, but 
that rule is not applicable to this entry. 26 Thus in the present case, adjectives such as 
necessary and possible will include as part of their sparse lexical entries the stipulation 
that the lexical rule that usually relates IT-SUBJ and S-SUBJ members does not apply 
to these entries. Calling that lexical rule LR-Intraposition, given in (39), the entry for 
necessary can then be represented as in (40). 
Rule LR-Intraposition lexical rule 
(39) 
LR-Intraposition 
LE2-Classes - IT-SUBJ = LE1-Classes - S-SUBJ 
26 For a more complete discussion of the types of exceptional behavior exhibited by lexical entries with 
respect to lexical rules in this framework, see Flickinger 1987 pp. 122ff. 
305 
Computational Linguistics Volume 18, Number 3 
(40) 
necessary-1 
Superclasses Adjective, It-Subj 
Spelling "necessary" 
Phonology /nEsIseri/ 
Lexical-Rules (LR-Intraposition Not-Applicable) 
It may be worth noting that it seems unlikely that properties such as the applica- 
bility of lexical rules can be incorporated into the feature system of HPSG (i.e., in any 
explanatory way). They seem inexpressible because they are a kind of second-order 
property. This is, in fact, exactly the sort of information that suggests to us that a 
lexicon may have to be more than a particular kind of feature system. See Pollard and 
Sag (1987:209, note) for a concurring view. Krieger and Nerbonne (1991), on the other 
hand, propose a feature-based treatment of lexical rules that allows the expression of 
exceptionality (without using rule applicability features)Y 
Appendix B: Refinements and Lexical Modifiability 
In addition to allowing extensions painlessly, we expect a lexical system to be easily 
modified. This is of practical value given the relatively inexact state of present linguistic 
knowledge. Linguistic descriptions are under frequent revision, and lexical systems 
must accommodate this. In the present section we examine several refinements of the 
analyses above as a means of demonstrating the modifiability of structured lexicons. 
We wish to underscore the richness of detail that demands accommodation even in 
this one corner of the lexicon, and we hope to probe the limits of the formalism we 
have adopted for this lexical representation. 
We began our analysis by presenting two variants of adjectives like easy, one with 
an expletive it subject, and one with a normal NP subject and a verbal complement 
containing a gap. There is, of course, a third variant for most adjectives of this kind, 
one with an infinitival VP or S as its subject, and no verbal complement, as illustrated 
in (41). 
(41) a. To talk to Bill would be great. 
b. For me to talk to Bill would be great. 
c. For Bill to lose this race would be great for Mary. 
This selection for infinitival subjects is a property shared with other classes of 
well-studied predicates, including verbs like bother and require, as illustrated in (42). 
(42) a. For me to talk to Bill would bother Mary. 
b. To win this race will require your fullest commitment. 
27 The initial implementation of this lexicon was reported in Flickinger et al. (1985), and was done in 
HP-RL, a language derived from MIT's frame representation language, FRL (Goldstein and Roberts 
1977). It has since undergone reimplementations in Common Lisp, Common Objects, and CLOS. The 
work reported on here was implemented and saw daily (experimental) use for over two years. The 
basic analyses in Section 4 were all implemented and thoroughly tested through a good variety of 
surrounding grammars and application efforts. We also suggest analyses in the main body of the paper 
that were not implemented fully, in particular in the section on adjectival specification (too and enough) 
and nondenoting nominals S. is a pleasure to see. 
306 
Dan Flickinger and John Nerbonne Inheritance and Complernentation 
To see that these infinitival subjects must be lexically licensed, consider the ex- 
amples in (43), where at least some members of the IT-SUBJ class cannot appear with 
such subjects. 
(43) a. *In the final analysis, to win this race will not be necessary. 
(cf. Winning this race will not be necessary.) 
b. *To talk to Bill is possible only in the mornings. 
We thus define a subclass of INCOMPLETE here named S-SUBJ, similar to IT-SUBJ 
and SLASH-COMP, to identify the relevant properties exhibited by lexical entries for 
the adjectives and verbs in (41-42). 
(44) 
S-SUBJ 
Superclasses 
Complements 
Subject-features 
Incomplete 
(Category Verb) (VForm Infinitival) 
(Complete +-) 
To this class belong verbs like require and bother, but we must also define a subclass 
that we call S-EASY for adjectives like great and difficult, since these, unlike the verbs, 
also permit an optional PP-for phrase, provided as before by the FOR-EXPERIENCER 
class. 
(45) 
S-EASY 
Superclasses S-Subj, For-Experiencer 
It appears that in general adjectives of the IT-EASY class alternate with entries like 
those in (41), while adjectives of the IT-SUBJ class do not. Hence we are tempted to 
define the lexical rule relating adjectives having it subjects with those having infinitival 
subjects so that the rule holds between the two classes S-EASY and IT-EASY. However, 
verbs like bother and require with entries that are members of the IT-SUBJ class should 
also be covered by this same lexical rule, suggesting that it must hold between the 
two classes IT-SUBJ and S-SUBJ. This leaves us the task of excluding those IT-SUBJ 
adjectives like necessary and possible that do not have S-SUBJ counterparts. 
Distinctions among Unbounded Dependencies 
A second refinement of our analysis of easy adjectives is motivated by examples like 
those in (46), which show that some further constraints need to be placed on the gappy 
verbal complement that such adjectives subcategorize for. Informally, the generaliza- 
tion seems to be that extraction is not possible out of finite clauses embedded within 
the complement to easy adjectives, but is otherwise licensed. 2s 
28 Jones (1990) attributes the observation to Chomsky (1977). 
307 
Computational Linguistics Volume 18, Number 3 
(46) a. Bill was easy to get Mary to hire. 
b. Palm trees are hard to learn to climb. 
c. Arias are fun to try to sing. 
d. *Bill was easy to see that Mary admired. 
(cf. Bill, it's easy to see that Mary admires.) 
e. *Palm trees are hard to learn that one can climb. 
f. *Arias are fun to insist that people sing. 
General constraints imposed by the framework we have adopted here prevent us 
from attempting to describe these facts by making easy adjectives select for a verbal 
complement whose head requires its complement to be nonfinite. 29 Instead, we follow 
Hukari and Levine (1991), who propose that two types of unbounded dependencies 
might be distinguished, with one dependency path, marked by the new binding feature 
SLASH ~, treating finite S's as islands, while the ordinary SLASH feature marks the 
usual dependency path that is insensitive to finite S nodes. Then easy adjectives of the 
SLASH-EASY class would more precisely subcategorize for an infinitival complement 
that has an NP gap of the marked SLASH' variety rather than the usual SLASH. 
To illustrate, consider example (46d): the SLASH' feature that would (by hypothesis) 
be introduced at the extraction site within the embedded S that Mary admired will 
be passed up from that site by a general principle, but will stop its ascent when it 
reaches that S. Assuming that the verb see does not select for this unusual kind of 
SLASH ~ complement, the sentence will not be admitted as grammatical. In contrast, 
the example in (46a) will still be admitted since the SLASH t introduced at the extraction 
site in to hire will be faithfully passed up by the same binding inheritance principle 
until it reaches the node dominating to get Mary to hire. Now this VP\[SLASH / NP\] 
is, according to our proposed refinement, precisely the kind of complement that easy 
requires, so the sentence is grammatical. 
Of course, to properly defend this addition of SLASH I to the collection of binding 
features for English, we would like to find independent evidence of the claim that 
finite S's can serve as islands for SLASH. We leave the matter here as one meriting 
further study, and refer the interested reader to Jacobson (1987) for a similar suggestion 
to distinguish various slash attributes. 
Pied Piping 
A third refinement of the analysis given above is motivated by examples like those in 
(47), where the easy adjective appears in a noun phrase with an infinitival complement 
containing a pied piping construction, not accounted for in what we have said thus 
far. 3° 
(47) a. Mary is an easy boss for whom to work. 
b. New York would be an awkward city from which to flee. 
c. Bill might be a hard person in whom to confide. 
The most straightforward characterization of phrases like for whom to work is to 
describe them as infinitival VPs containing a relative pronoun. Since English indepen- 
29 For discussion, see Flickinger 1987, pp. 67ff; Pollard and Sag 1987, p. 143. 30 Chomsky (1977) cites examples like these; we appreciate Anthony Kroch's bringing them to our 
attention. 
308 
Dan Flickinger and John Nerbonne Inheritance and Complementation 
dently prohibits relative pronouns from appearing in situ within a verb phrase, the 
only way such a phrase can be produced is to have a pied piped prepositional phrase 
extracted from the VP and sister to it. 31 Assuming that such a phrase must be admit- 
ted by the grammar, we can formally represent its syntactic category as shown in the 
following definition for the new class REL-EASY. What we make explicit here is the 
idea that adjectives of the SLASH-EASY class have corresponding members (linked 
via a lexical rule similar to ones seen above) that take an unusual kind of VP com- 
plement, differing further in that these REL-EASY adjectives do not seem to license 
an optional PP-For complement. The final property identified in this class is that its 
members are marked as not predicative, effectively restricting its members to attribu- 
tive adjectives. This property accounts for the ungrammaticality of the examples given 
in (49). 
(48) 
REL-EASY 
Superclasses 
Features 
Complements 
XComp-features 
Control,Adjective 
(Predicative -) 
(VForm Infinitival) 
(REL (Category Noun) (Complete +) 
(NForm Normal) (Predicative -) 
(Case Accusative Dative) ) 
(49) a. *Bill is easy for whom to work. b. *Bill is a pleasure for whom to work. 
In both examples in (49), the complement of the copula is must be predicative, 
but the phrases headed by easy and pleasure would have to be nonpredicative in or- 
der for those heads to license their VP\[REL NP\] complements. Indeed, (49b) is also 
ruled out by another constraint that we finessed in our brief introduction of pleasure 
nouns, for clarity of exposition: we described such nouns as belonging to the ordinary 
COMMON-NOUN class, but that assignment also needs refining in a more detailed 
account, since pleasure nouns exhibit only a few of the properties of regular common 
nouns. In particular, it seems clear that these nouns must be predicative, and hence 
unfit for membership in the REL-EASY class, since attempting to assign them to this 
class would introduce a conflict of inherited values for the attribute PREDICATIVE, 
and such conflicts are prohibited, as we noted above in our introduction to the general 
framework. 
31 Among other details, the phrase structure linking rule that would be necessary to admit this 
\[VP\[REL NP\] --* PP\[REL NP\],VP/NP\] construction will also have to be made explicit in a fuller 
analysis than we provide here, but that should not be problematic. We have in mind a simple 
generalization of the sentential linking rule, so that no novel rule would be required. 
309 

