A Computational Approach to Binding Theory* 
Alessandra Giorgi, Fabio Pianesi, Giorgio Satta** 
Istituto per la Ricerca Scientifica e Tecnologica, 38050 Povo (Trento), Italy 
e-mail: satta@irst.it ; satta@irst.uucp 
Abstract 
This paper is a first step towards a 
computational account of Binding Theory (BT). 
Two algorithms that compute, respectively, 
Principle A and B have been provided. Particular 
attention has been devoted to possible 
interactions of BT with other modules of the 
linguistic theory, such as those ruling 
argumental chains. Finally, the computational 
complexity of the algorithms has been studied. 
1 Introduction 
This work is a contribution to the computational study 
of the referential properties of Noun Phrases (NPs). In 
particular, it focuses on the disjoint reference constraint 
for pronouns, and on the binding requirement for 
anaphors. 1 Unlike other attthors, we do not output actual 
references for pronouns. 2 The reasons for such a move 
will be discussed below. 
In pursuing these goals, we will refer to Binding 
Theory (henceforth BT), as developed within the 
Government and Binding framework by Chomsky and 
his collaborators (see Chomsky, 1981, 1986); in 
particular, algorithms will be presented that compute 
Principles A and B of BT. 
Section 2 presents a brief introduction to BT. In 
Section 3, we will introduce a formal (computational) 
apparatus, which will then be used to formulate the 
algorithms. Finally, some considerations about their 
formal properties will be discussed. Section 4 
illustrates, by means of an example, how the algorithms 
work. Finally, in Section 5, our approach and results 
will be compared with those already present in the 
literature. 
*Every part of this work has been elaborated jointly by the three 
authors. However, as far as legal requirements are concerned, 
A.Giorgi takes responsability of Sections 1 and 2, F.Pianesi of 
Sections 4 and 5 and G.Satta of Section 3. 
**This work has been done while G.Satta was completing his 
Doctoral Dissertation at the University of Padova (Italy). 
1This work is part of a larger one providing a computational 
account of Binding Theory. See Giorgi, Pianesi, Satta (1989a), in 
which the present approach is extended to principle C of Binding 
Theory and weak-crossover core cases, and Giorgi, Pianesi, Satta 
(1989b), where the general problems of Binding Theory 
verification and satisfiability are addressed. 
2See Berwick and Weinberg (1984), Correa (1988), htgria and 
Stallard (1989), Berwick (1989); for a different approach see Ristad 
(1989). 
2 Introduction to Binding Theory 
Binding Theory (BT) is a module of the Government 
and Binding theory ruling the distribution and the 
referential properties of anaphors (such as himself), 
pronouns (such as him and his) and R(eferential)- 
expressions (such as John, the man I met yesterday, my 
sister, etc). Here we will briefly illustrate its scope, 
without entering into a detailed analysis. 
It is a well known fact that lexical items, such as 
Noun Phrases, must undergo an interpretation process 
by which they are assigned a referent. Such a process is 
ruled by principles that vary according to the nature of 
the item in question, i.e. anaphor, pronoun or R- 
expression. A first generalization can be stated as 
follows: anaphors must have an antecedent in the 
syntax, i.e. in the same sentence where they appear; 
pronouns can directly identify a referent in the world or 
in the previous discotu'se; R-expressions are intrinsically 
referential, i.e. they need no antecedent? Consider the 
following examples: 
(1)a. Johni loves himself i 
b. *I love himself 
In (1)a. the anaphor himself takes John as an antecedent, 
i.e. in technical terms, it is bound by it; irt (1)b, for 
morphological reasons, I cannot work as an antecedent 
for himself, so that the whole sentence is ruled out. 4 
Consider now what happens in the case of pronouns: 
(2) John thinks that Mary likes him 
him can either refer to John or to someone else in the 
world, for instance to someone mentioned in the 
previous discourse. The conclusion up to this point can 
be summarized as follows: an anaphor must have an 
antecedent, a pronoun can have one, an R-expression 
cannot. However further properties must be taken into 
account; let us consider pronouns again: 
(3) Johni likes him,i 
ttim cannot be interpreted as John, contrary to what 
happens in (2): there is a "negative" condition on 
3For reasons of space, the referential properties of quantified 
expressions and those of the so-called epithets are not considered 
here. 
4R-expressions cannot be coindexed with any c-commanding 
item (see below for a definition of c-command). Consider the 
following example: 
(i) John likes John 
In this case, given that the first R-expression c-commands the 
second one, the two occurrences of John must refer to different 
persons. 
120 l 
pronouns, since they cannot have an antecedent lying 
within a certain contexl:. Technically speaking, a 
pronoun must be free in its local domain; a similar 
locality condition holds with respect to anaphors, since 
~tll anaphor must have an antecedent inside a certain 
domain: 
(.4)a. *John i thinks that Mary likes himselfi 
b. Mary thinks that John i likes himselfi 
(4)a. is ungrammatical because the intended antecedent, 
John, lies too far away. 
The local domain is the so-called Complete 
Functional Complex (CFC) pertaining to the item in 
question, i.e. containing both the item itself and its 
governor: 5 
(5) ?,is a Complete Functional Complex iff one of 
the lollowing holds: 
a) y is the (minimal) domain in which all the 
0-roles pertaining to a lexical head m'e realized; 
b) y is the (minimal) domain in which all the 
grammatical functions pertaining to that head 
m'e realized. 
Finally, an important structural condition must hold, 
namely c-comm~md: (of. Chomsky, 1981; 1986): 
(6) cx c-commands fi iff 0'~\]3, a does not dominate 
\]3 and the first node, y, dominating cx also 
dominates \]3. 
"1-'o be bound by a given item actually means: to be 
coindexed and c-commanded by that item; whereas to be 
free means: not to be coindexed with a c-cormnanding 
item (non c-commanding items might work, on the 
olher hand, as possible antecedents). The principles of 
b:iuding can be expressed as follows: 
(7) A: An anaphor is bound in its local domain 
B: A pronoun is free in its local domain 
C: An R-expression is flee 
Note that, as will be shown below, in our system we 
can also handle constructions involving the so-called 
pro-drop phenomena, found in languages like Italian, 
Spanish and so on. Consider the following Italian 
ex ample: 6 
(8) proi arriva lui i 
lit: at:rives he 
The system must know that the coindexation in (8), i.e. 
the fact that the pronoun lui is coindexed with a c- 
commanding empty category (the expletive pro), is not 
a Principle B violation. 
3 The Algorithms 
With respect to the two principles considered here, i.e. 
Principles A and B, the output of BT can be represented 
"hi'he definition in the text is Ihe one foand in Giorgi (1987); cfr 
also Chomsky (1986). Note that, in most cases, the CFC coincides 
wilh the first Senience or Noun Phrase dominating the ilem in 
qudstion. However, this is not always the case and the systems 
defining the binding domain this way often nm into trouble; this 
point will be fnrther considc~cd in Seclion 5; sue also Giorgi, 
Plainest, Satta (1989a). 
61npro-drop languages (sec Chomsky 1981) typically, lhe 
subieet can be noJl lexica\[, i.e. can be an empty calegory, or can bc 
expressed postvel%ally, leaving an cxpletive empty category in 
subject position. 
as a lomml language Lv. More precisely, given a 
sentence w, let T be the set of all tuples t=<z'w, (x, fit, 
.... /7,~>, n_>0, where "cw is a parse tree for w, a is either 
an anaphor or a pronoun and the components fll ... fin 
represent any set of NPs. v Let us define Le c T to be 
the set of all tuples such that the following conditions 
hold: 
(i) if o: is an anaphor, \]31...fin are all and only the 
items that can be antecedents for c~, according to 
Principle A of BT; 
(it) if c~ is a pronoun, then ill...fin are all and only 
the items disjoint from ~x, according to Principle B 
of BT. 
The algorithms to be presented can be seen as 
recogniscrs for L s. 
3.1 Definitions 
Let N:(nl ..... nq}, q>_l, be the set of all nodes in rw. 
We will also assume the following functions and 
predicates: s a function father from N to Nu {_t_}; a 
function siblings from N to '/'(30; a binary predicate 
agreement, defined on NxN, such that agreement(n1, 
n2)=TRUE iff the agreement features of n 1 and n2 are 
mutually compatible; a unary predicate pt-antecedent, 
defined in -N, such that pt-antecedent(n)=TRUE iff n is 
a maximal projection of a head N o (a Noun) within T~. 
Definition 1 A binary predicate domain is defined in 
N×N in such a way that domain(ha, hal=TRUE if f: 
(i) n,~ is the least constituent such that either all the 
O-roles pertaining to a lexical head are realized, or all 
the grammatical functions pertaining to the same 
lexical head are realized; 
(it) ny is the lexical governor of na and 
filther(n ~)~filther(n,4).9 0 
Condition (it) has been explicitely introduced in order to 
lake care of cases of government across the boundaries 
(see Section 3.3). 
To account for the interaction of BT with pro-drop (cf. 
ex.(8) above), we also need the following definitions. 
Let ch-mark be a procedure defined on Nx&(: whenever 
ch-mark(n,nm) is invoked, if n is a landing-site of a 
chain ~° c within rw, every node nc such that nc belongs 
to c, gets marked with a distinctive mm'ker, which will 
be assumed to be the second argument node n m. This 
marking relation will hold until a new call to the 
procedure takes place for any node corresponding to the 
same chain as n, with a different node-marker. We need 
7We assume lhat all the principles of the theory have already 
been applied to tile sentence. Such an asstimption is reasonable, 
given the modular nature of tile theory; see Chorasky (1981, 
1986). 
8In tim following, the symbol .L is meant to denote the undefined 
element. Also, for a generic set .~l, ~v(51) denotes the set of all 
t×)ssible subsets of .a (the power set of A). 
9The exact meaning of linguistic notions such as: 0-roles, 
functional-.roles, lexical governor and government can be found in 
Chomsky (1981, 1986). A computaliona\] account of fl~e no,.iov, of 
\]ocal domain can be found in Giorgi, Pianesi and Satta (1989a), 
along with some tbrmal properties of the predica'.e domain. 
10l(oughly, tim notion of chain can be defined (cf. Chomsky, 
1986) as the set of coindexed positions (landing sites) pertaining 
to the same syntactic object (uhetv only one of sLicl't posili'.,ns is 
lexically filled). 
2 121 
also a function ch-marker, from 5% to Nvo (±\], defined 
such that ch-marker(n)=nm iff n is a landing-site of a 
chain c within ~,,, and a previous call to the procedure 
ch-mark has m~ked each node in c with the marker-node 
n m . 
3.2 Algorithm Schemata 
The two algorithms behave in a very similar way; they 
take as input a node in N corresponding to an NP in 
Vw, and analyse some specific relations between the 
input node and each node in N that c-commands the 
input node, up to certain specific domain. The c- 
commanding relation is implicitly encoded in the way in 
which the algorithms apply the two flmctionsfather and 
siblings. 
An Algorithm for Principle A 
Given an input node n which corresponds to an 
anaphor in l:w, the algorithm outputs a list of nodes 
corresponding to "actual antecedents" for the anaphor 
itself. The algorithm looks for a "potential antecedent" 
of the input anaphor, starting from node n and 
proceeding from bottom-to-top. As soon as a potential 
antecedent is found, the algorithm restricts its search to 
the local domain it has just identified. Note that each 
potential antecedent must pass the agreement check to 
be considered an actual antecedent n 
We also consider some cases of referential circularity; 
in particular, problems arising in pro-drop 
constructions. More specifically, a node which belongs 
to the chain also containing the anaphor, cannot be 
collected as a potential antecedent. The following 
circularity check is therefore included: every chain whose 
landing-sites dominate the input anaphor, up to the 
domain of interest, is marked by the procedure ch-mark 
using the input-node as a marker. In this way a node c- 
comnmnding the input node and corresponding to a 
landing-site of a chain marked by the latter, cannot be 
taken as a potential antecedent for the input-node itself 
(for more discussion, see Section 5). The same 
mechanism also ensures that, for every possible chain, 
only one of its landing-sites is ever considered as a 
potential matezedent. 
Algorithm 1 
input-node: A node corresponding to an anaphor in 'r~. 
Output: A list of nodes in N corresponding to actual 
antecedents for the input anaphor. 
Method. 
Step 1: Let input-node be the value of the program 
variable present-node, hfitialize also the program 
variable local .domain-flag to the value FALSE and 
invoke the procedure ch-mark(present-node, input- node). 
Step 2: For each value of the program variable present- 
sibling in siblings(t)resent-node), if ch-marker(present- 
sibling)veinput-node and pt-antecedent(present- 
sibling)=TRUE, perform the following actions. Set 
the program variable local-domain-flag to TRUE if it 
is FALSE and invoke the procedure ch-mark(present- 
l lAccording to Chomsky (1986), the existence of the potential 
anlecedent for an anaphor is crucial in defining its local domain. 
Note timt such an item is not necessarily the actual antecedenl, 
sibling, input-node); furthermore, if 
agreement(present-sibling, input-node)=TRUE then 
output present-sibling. 
Step 3: If father(present-node)=L, go to Step 4, 
otherwise let father(present-node) be assigned as the 
value of present-node. Invoke the procedure ch- 
mark(present-node, input-node). If local-domain- 
flag=FALSE then restart at Step 2. Otherwise there 
are two possibilities: if domain(present-node, input- 
node )=FALSE then restart at Step 2; if 
domain(present-node, input-node)=TRUE go to Step 
4. 
Step 4: Stop. ca 
An Algorithm for Principle B 
The algorithm starts from an input node that 
corresponds to a pronoun in rw. The algorithm visits all 
nodes in Nwhich correspond to elements c-commanding 
the input pronoun and lie inside the local domain; 
finally, it outputs a list of disjoint elements. Indee(1 the 
algorithm is procedurally very similar to the one given 
for Principle A, with minor changes due to the 
differences in the definitions of the local domain. 
Algorithm 2 considers each chain only once, as does 
Algorithm 1. Observe that if a pronoun belongs to a 
certain chain, it cannot be disjoint from other elements 
of the same chain. An identity check is then carried out 
by the algorithm in the following way: the chain, which 
the input-node belongs to, is mitiatly marked by the 
procedure ch-mark. Then, every node that c-commands 
the input-node inside its local domain, corresponding to 
a landing-site of this marked chain, will not be inserted 
in the output list of Algorithm 2 (see Section 5). The 
details are the lbllowing: 
Algorithm 2 
Input-node: A node corresponding to a pronoun in r,. 
Output: A list of nodes in N corresponding to the 
disjoint elements for the input pronoun. 
Method° 
Step 1: Let input-node be the value of the program 
variable present-node. Invoke the procedure chain- 
mark(present-node, input-node). 
Step 2: For each value of the program variable present- 
sibling in siblings(present-node), perform the 
following action. If ch-marker(present-sibling)~-input- 
node and pt-antecedent(present-sibling)=TRUE, output 
present-sibling and invoke the procedure ch- 
mark(present-sibling, input-node). 
Step 3: Iffather(present-node)=Z, go to Step 4, 
otherwise let father(present-node) be assigned as the 
value of present-node. If domain(present-node, input- 
node)=FALSE then restart at Step 2, otherwise go to 
Step 4. 
Step 4: Stop. 
3.3 Some Formal Results 
Some properties of Algorithms 1 and 2 will be stated; 
see also Giorgi, Pianesi and Satin (198%). 
Theorem 1 The predicate domain(present..node, input- 
node) holds true at Step 3 in Algorithms 1 and 2 iff 
present-node corresponds in rw to the minimal CFC 
containing both input-node and n~, where n 7 is the 
lexical governor of input-node. 
122 3 
Proof 
'Only if Condition (i) in Definition 1 guaranties that 
present-node is a CFC, as defined in (5). Furthermore 
presenbnode dominates input-node at Step 3, as it is 
easy to show. It remains to demonstrate that present- 
node dominates n.¢. A government relation between n 7 
and input-node can only be attained within the following 
three structural configurations. In the first, government 
is realize.d under sisterhood; thus, every node that 
dominates the governee will also dominate the governor. 
In the second configuration the govcrnee is attached 
higher than its governor, within the maximal projection 
of the latter; again, every node that dominates the former 
will also dominate the latter. The third possibility 
concerns the so called government across boundaries: 
when a maximal projection ZP (or a Small Clause) is in 
sisterhood relation with a lexical category X °, then the 
latter can govern the specifier position of Ihe former (or 
the subjex:t position, in the case of a Small Clause). ZP 
may well be a CFC, in the sense of (5), but it does not 
contain the governor X °. Condition (it) in Definition 1 
explicitely rules out this case, so the claim is proved. 
'If' The proof immediately follows from the given 
analysis of the possible configurations of government 
between the nodes n 7 and input-node, and from 
Definition 1. 
XP /)', 
X ° =n yp = input-node y 
Figure 1 
XP 
yp = input-node 
X 0 = n 7 ... 
Figure 2 
Theorem 2 Let "Cw be an X-.bar description for some 
sentence w such that all the principles of GB hold trtte 
for it, apart from the BT principles, and let N be the set 
(~f all nodes ill "~w. When input-node is assigned a value, 
which corresponds to an anaphor a in "rw, Algorithm 1 
computes the whole list of nodes in N that corresponds 
to the antecedents of a, in the sense of Principle A of 
BT. 
f:'roof Omitted. La 
Theorem 3 Let "Cw and N be as in Theorem 2. Given 
as input a node that corresponds to a pronoun ~ in "cw, 
Algorithm 2 computes the whole list of nodes in N 
that must be disjoint from a, in the sense of Principle 
B of BT. 
Proof Omitted. c) 
Questions about time complexity are now addressed for 
Algorithms 1 and 2 (we assume, as the reference model 
for comi)utation, a RAM). 
X* 
YP = input-node Z' 
Figure 3 
Theorem 4 The running times of Algorithms 1 and 2 
are given by two functions fAz andfA2, such thatf4 ~, 
fa2e O(n), where n is the length of the sentence under 
analysis. ',2 
Proof (outline) From elementary considerations about 
X-bar Theory, ~3 it can be argued that set N has .'dze 
bounded by an expression of the form cxn+c2, It is easy 
to show that no node in N is visited more than once by 
Algorithms l and 2 and that a constant amount of time 
is spent in visiting each node; then the result follows. 
4 A running example 
Let us see how Algorithm 1 works with the following 
sentence, giving, as input node, the one corresponding 
to the anaphor herself: 
(9) lIP Mary \[r \[vt' \[v' sees herself \[pp in the 
minvr\]\]\]\]\] 14 
At the beginning, the variable present-node is set to the 
value of input-node, i.e. the anaphor node, while the 
variable local-domain-flag is set to FALSE. Then 
Algorithm 1 enters Step 2, where it scans the anaphor's 
siblings. Once present-sibling is set to the PP node, pt- 
antecedentQ)resent-sibling)=FALSE. At this point, 
Algorithm 1 exits Step 2 and enters Step 3, where 
present-node is set to the value of father(present-node), 
in this case, the V' node. Given that local-domain- 
.flag=FALSE, Algorithm 1 enters Step 2 again. We do 
not follow the whole computation but directly skip to 
the point where Algorithm 1 enters Step 2 with present- 
12The standard notation g(n)c Off(n)) means that there exist a 
positive constant M and an integer n o such that g(~,)<Mf(n) for all 
?~?t 0. 
13See J ackendoff (l 977). 
ldAnother possible analysis of this sentence hypothesises, as 
the argument of see, a Small Chmse inchlding herself and the 
predicate PP. 
4 123 
node=I'. The only sibling of I' is the subject NP and, 
setting present-sibling to it, one has that pt- 
antecedent(present-sibling)=TRUE. Algorithm 1 then 
sets local-domain-flag to TRUE; furthermore, 
agreement(present-sibling, present-node)=TRUE, so that 
the value of present-sibling is output. After that, 
Algorithm 1 enters Step 3 and sets present-node to 
father(present-node), i.e. to the topmost (IP) node. Now 
local-domain-flag=TRUE and domain(present-node, 
input-node)=TRUE (i.e. IP is the local domain for the 
input anaphor); therefore Algorithm 1 enters Step 4 and 
then stops. 
5 Discussion 
By fixing an upper bound, it has been shown that the 
computational complexity of the recognition problem of 
a language that encodes Principles A and B of BT is in 
p.~S These results are similar to those obtained by all 
authors who have studied BT from a computational 
point of view (Correa, 1988; Ingria and Sta\[lard, 1989; 
Berwick and Weinberg, 1984; Berwick, 1989). 
Nevertheless, with respect to such works we have both 
taken a rather different perspective and paid more 
attention to the subtleties of the linguistic theory. 
Previous works were mainly concerned with providing 
actual referents (actual indexations) for the NPs of a 
sentence. We claim, on the contrary, that Principles A, 
B and C per se are not sufficient for this puq~ose, since 
BT only restricls the search space for indices selection, 
and does not actually provide them. For instance, Correa 
(1988) proposes an algorithm that builds lists of 
antecedents for pronouns and anaphors, and 
complements it with a Binding Rule, that selects, for 
each item, an indexation from such lists. However, the 
selection of an antecedent for a certain item could affect 
the indexation of other nodes, leading to violations of 
Principle B. 16 In tb.is framework, a related problem 
arises considering split antecedents for pronouns; in fact, 
a pronoun can be coindexed with a set of items, provided 
that each of them has a different thematic roleJ 7 This 
point has never been explicitely addressed in 
computational works; nevertheless, if the purpose is to 
output actual indexations, it seems to us that the only 
possibility, in order to consider split antecedents, would 
be to compute lists of possible antecedents for pronouns 
and then to consider their power set; this way, however, 
the search space becomes exponentially large. 
Furthermore, the interactions of the referential properties 
15¢o denotes the class of languages recognizable in polynomial 
time by a deterministic Turing machine. 
16Consider, for instance, the following sentence: 
(i) Mary says that she saw her 
in this case both embedded pronouns can take Mary as an 
antecedent, according to Correa's Binding Rule, leading to a 
violation of Principle B. In our opinion, to avoid this incorrect 
result, it is necessary to put together the constraints that have been 
separately computed for each item according to Principles A and B 
(and C); this way we can account for the interactions between 
coindexations and disjointness. A possible way to do it, is to pose 
the problem of BT verification, i.e. whether a given index 
assignment for the NPs of a sentence complies with the 
restrictions of BT. See Giorgi, Pianesi, Satta (1989b). 
17For instance, in Mary tom John that they should go home the 
pronoun they can refer to the complex antecedent constituted by 
Mary plus John. 
of a split antecedent with those of other items (,possibly 
other split antecedents) would thereby hardly be 
addressable. TM 
Finally, given the referential properties of pronouns, 
it seems that there is no point in trying to use the 
grammatical knowledge of BT to hypothesise 
intrasentential antecedenks, t9 
Another crucial aspect concerns the treatment of local 
domains, whose importance has often been 
misconsidered in computational works on BT. Such a 
notion has been mainly seen as a static one, whereas, in 
our interpretation, the value of the actual domain 
depends on the interaction of structural and lexical 
properties of at least two different positions in the 
derivation tree. For example, in Ingria and Stallard 
(1989), an S node is taken to be the binding domain for 
every node it dominates. Consider, however, the cases in 
which government of the specifier position of a 
maximal projection is obtained through an external 
head; this situation arises, for instance, in exceptional 
case marking examples, as in John believed him to be 
intelligent. Ingria and Stallard's static definition of local 
domain would lead to the conclusion that the pronoun, 
being dominated by an S type node (the embedded 
sentence), is free in that category and, thus, could be 
coindexed outside, for instance with the R-expression 
John. But this is ungrammatical; according to the 
definition adopted here, the domain of binding for him is 
the matrix clause, so that the pronoun must, correctly, 
be free in it, i.e, disjoint from John. Our approach also 
improves on Ingria and Smllard's treatment of NP as a 
binding domain. If a node NP containes a possessive 
then they consider it a binding domain for all the nodes 
it dominates, except the possessive itself. There are at 
least two problems, though. First, they do not predict 
that a pronoun subcategorized for by the head cannot be 
bound in the domain of the NP; second the well-known 
not complementary distribution of pronouns and 
anaphors in the specifier position of an NP cannot be 
accounted for. The definition of binding domain adopted 
in (5) and the way it is computezt allow our algorithms 
to avoid these problems; see Giorgi, Pianesi, Satta 
(1989a). 
As a final remark, note that in this work the 
interaction with A-chains has been explicitely 
considered. This problem is particularly important in 
Italian which, being a pro-drop language, admits 
sentences like (8) and (10): 
(10) Giannii ha detto che proj arriver~t \[la propriai 
madre\]j 
lit.: Gianni told that will arrive self's 
mother 
Gianni told that his mother will arrive 
18Also the so called weak crossover phenomena may raise some 
problems. Roughly, pronouns cannot be coindexed with non c- 
commanding quantified expressions, as in *His imother loves 
\[every boY\]i , where the embedded pronoun cannot be taken to refer 
to the quantified expression. But this fact raises some problems for 
both Correa's and Ingria's approach. 
19pronouns can refer intersententially or deictically; note that 
this property is shared with certain R-expressions, like the 
epithets, which obey Principle C (see Ha'ik, 1984). 
124 5 
in (8) the postverbal slibject pronoun is coindexed with 
the expletive pro, but a procedure looking for disjoint 
elements would output a list containing pro (it lies in 
the local domain of the pronoun) thereby giving rise to 
a contradiction: the pronoun is coindexed with the 
expletive pro but must be disjoint fi'om it. In (10), we 
must avoid the anaphoric possessive proprio being 
coindexed with the c-commanding expletive pro, in order 
to rule out circular interpretations. The circularity and 
identity checks, discussed in Section 3.2, explicitly take 
care of these cases. 

References 

Berwick R.,(1989) , Natural Language Computational 
Complexity and Generative Capacity, to appear on 
Computers and Artificial Intelligence 

Berwick R. and Weinberg A., (1984), The Grammatical 
Basis of Linguistic Pe(\[ormance, MIT Press, 
Cambridge MA 

Chomsky N., (1981), Lectures o~l Government and 
BindMg, Foris, Dordrecht 

Chomsky N., (1986), Knowledge of Language, Praeger, 
New York 

Correa N., (1988), A binding Ru!e for Government- 
binding Parsing, Proceedirlgs of COLING, Budapest 

Giorgi A., (1987), The notion of Complete Functional 
Complex: Some Evidence fi'om Italian, Lb~guistic 
Inquiry, 18,3 

i Giorgi A., Piancsi F., Satta G.,(1989a), Towards a 
Complexity Analysis of Bindiz:g Theory, Technical 
Report n.8911-06, IRST, Trcnto, It. 

Giorgi A., Piancsi F., Satta G.,(1989b), The 
Compat~tional Conlpiexity of Binding Theory's 
Satisfiability and Verification, ms, IRST, Trento, It 

l ia?k I., (1984), Indirect Binding, LiJTguistic l, zquiry, 15, 

\]ngria R.P.J., Stallard 1),(1989), A Computational 
Mechanism for Pronominal Reference, Proceedings 
of the 271h Anmml Meeting of the Association for 
Comptltational Linguistics, Vancouver ACL

Jackendoff R., (1977), X-bar 5)ntax." A study of l)hrase 
Structure, MIT Press, Cambridge MA 
