Law in the Information Society: 5th International Conference of the Institute of Legal Documentation (IDG) of the Italian
National Research Council, December 2-5, 1998, Florence.
Legal Contract Drafting at the Micro-Level
Aspassia Daskalopulu
Dept of Information Systems and Computing
Brunel University, Uxbridge Middlesex UB8 3PH, UK.
E-mail:  Aspassia.Daskalopulu@brunel.ac.uk
Abstract
This paper reports on research whose general aim has been to explore the potential for developing logic-based
tools for the analysis and representation of legal contracts [Daskalopulu 1998].  The intended applications of
such representation include both the drafting of new contracts and the management and administration of
existing ones, that is to say, the general problem of storing and retrieving information from large contractual
documents, and more specific tasks such as monitoring compliance or establishing parties’ duties/rights under a
given agreement when it is in force.  Experimental material is drawn mostly from engineering contracts, which
are typically large and complex and contain a multitude of interrelated provisions.
The term ‘contract’ is commonly used to refer both to a legally binding  agreement between (usually) two parties
and to the  document, that records such an agreement.  A previous paper [Daskalopulu & Sergot 1995] was
concerned with  documents and the representation of contracts at the  macro-level : the emphasis is on issues
relevant to the design of structurally coherent documents.  This paper deals with the contents of  agreements  and
representations at the micro-level.  Micro-level drafting is the source of a host of issues ranging from
representing the detailed wording of individual sections, to representing the nature of provisions (obligations,
powers, reparations, procedures), to representing their "fitness" or effectiveness in securing some party's best
interests.  The implications of these issues for knowledge elicitation and representation were discussed in
[Daskalopulu & Sergot 1997].  Various techniques are available to assist in aspects of this task, such as
disambiguating contractual provisions and in detecting inconsistency or incompleteness (for example [Allen &
Saxon 1993], [Alchourrón & Bulygin 1971], [Jones & Sergot 1993; 1996]).  The paper discusses a number of
such techniques and seeks to identify the limits on their applicability in the domain of legal contracts.
1   Introduction
This paper reports on aspects of a broader research project that seeks to develop logic-based tools for
the drafting of legal contracts, their management and administration.   ‘ Drafting’  refers to the problem
of creating new contracts, ‘ management’  refers to the general problem of storing and retrieving
related documents or parts of them and  ‘ administration ’  refers to more specialised tasks, such as
monitoring the compliance of contracting parties with a given agreement, or advising them on the
effects of individual provisions and their applicability.
In the context of this research emphasis has been placed on the drafting process: developing
representations of contracts for this task provides a main generic framework on which specialised
tools for management and administration can be attached as extensions that manipulate such
representations.  The importance of drafting is particularly evident in the case of engineering
contracts and long-term trading agreements.  This is because such contracts are typically large, often
running to hundreds of pages and accompanied by technical drawings and special schedules.  As
such agreements are long-term, a large number of issues need to be addressed in the corresponding
documents.  Consequently, the documents contain a multitude of interrelated provisions.  The
drafting of a new contract is a lengthy process where several drafts are created and negotiated upon.
As English Contract Law holds that written agreements contain all and only that which the parties
2
have agreed upon (the parole evidence rule, cf. [Atiyah 1989]), it is important not to make mistakes
or omissions as these might yield significant financial consequences should litigation arise.
The term  ‘ contract’ is commonly used to refer both to a legally binding agreement between (usually)
two parties and to the  document  that is used to record such agreement when the latter is put in
writing.  This research draws on this distinction and separates the problem of designing documents
from the problem of designing agreements.  Earlier work on the drafting of legislation [Bench-Capon
1987] drew a distinction between the formulation of policies and their implementation via legislative
text.  An analogy can be made whereby the formulation of a policy is similar to the formulation of a
specification (in software engineering terms) and is carried out by Parliament or the appropriate
policy-making body.  The drafting of legislation, extending this analogy, is then akin to
programming; the drafted legislation is intended to implement the specification —the policy that was
formulated.  A similar analogy can be made for the case of contracts.  The agreement between two
parties is negotiated and finalised as a specification of the goals that the business exchange it
regulates is meant to achieve.  The document that records it is the implementation of such
specification.  When developing tools for contract drafting, however, it is difficult to demarcate
precisely what is relevant to the design of documents from what is relevant to the design of
agreements.  Usually the people involved in the negotiation of agreements are also often responsible
for the production of the documents.  Moreover these two stages in contract formation are often
interrelated: what is agreed between the parties is what must be implemented in the document, yet it
is often the document  itself  that is being negotiated, for example what aspects of the business
exchange it will refer to, in what level of detail and often in what form of particular wording.  By
drawing explicitly on the distinction between documents and agreements, however, some aspects of
this process are facilitated, especially in cases where the creation of new contracts is based on
previous examples or model forms.  Hence, contract drafting can be discussed at two levels:
(i)   At the macro-level, the emphasis is in producing structurally coherent  documents that
implement an agreed specification.
(ii)  At the micro-level, the emphasis is in producing  well-formed  documents that implement an
agreed specification and indicate whether the agreed specification is well formed.
A previous paper [Daskalopulu & Sergot 1995] discussed contract drafting at the macro-level.  There
the process of constructing a new document, in the presence of existing previous examples or model
forms, is regarded as similar to computer-aided design subject to constraints.  Drafters use pre-prepared blocks of text in order to construct a new document of a particular kind, much in the same
way that a graphics designer uses geometric shapes to construct a picture.  This approach adopted the
idea that a document can be assembled from its components (cf. [Fiedler 1985], [Gordon 1992]) but
extended it further: the assembly process is subject to a collection of explicitly stated constraints that
govern the structure of documents.  These constraints express relationships between document
components and their satisfaction ensures that the output of the assembly process is a structurally
coherent document.  A computational model of this approach was produced where the structure and
interrelationships between constituent parts of documents were represented rather than the text of the
3
document itself.  That is, the representation of contractual documents at the macro-level does not
capture information about what a contract actually says  or  prescribes (what it  ‘ means’ ).
This paper is concerned with contract drafting at the micro-level, where the issue of interest is the
formulation of individual provisions ensuring that the overall document is well formed.  Section 2
considers criteria for well-formed individual provisions and documents.  Section 3 discusses a
number of techniques that are available to address some of these criteria.  It should be noted at this
point that in the course of this research an important observation came to light.  Contrary to
expectations, the sample contractual documents that were examined do not contain  all of the content
of the corresponding agreement.  This is because where a legislative Act exists, which regulates
trading agreements of some type, the provisions of the Act are  implied  into agreements of that type
(for example sales contracts conform with the Sale of Goods Act).  To attempt therefore complete
representation of contractual content we would need to represent domain information and relevant
legislation.  In this research, the focus has been on the representation of agreements as these emerge
from an examination of their associated documents.  The resulting representation is therefore not
complete in the sense mentioned earlier.
2   Well-formed contractual provisions
At the micro-level, the drafter’ s quest is for  well-formed  documents.  A question that arises naturally
is what makes a contractual document well formed.  The obvious but rather vague answer is that the
document should accurately implement the agreement on which the parties negotiated.  A whole host
of issues emerge from this, such as what makes a particular provision  ‘ good ’  for a party in terms of
the party ’ s expectations from the agreement and its broader goals, what implications does a provision
entail and so on.  In [Daskalopulu & Sergot 1997] some of these issues were discussed in more detail
with suggestions of the areas of research that might prove useful in addressing them.  This paper
concentrates on issues of consistency and completeness, but touches upon the roles that provisions
serve.
The sample contracts that were examined during the course of this research highlighted the following
roles that provisions play:
(i)   Descriptive: Such provisions provide definitions and guidance as to how certain terms are to
be interpreted and used in the context of the contract.  The term “Day” for example may be
defined to mean some period not necessarily comprising 24 hours.
(ii)  Prescriptive :  Such provisions determine certain behaviour for the parties by specifying
actions that must/may/must not be taken or states of affairs that must/may/must not obtain.
Usually such prescriptions are accompanied by a specification of circumstances or time periods
where the required behaviour is to take place.
(iii)   Procedural:  Such provisions specify procedures that need to be followed by parties in order
for certain states of affairs to be established.  Aspects of an agreement that are usually
4
accompanied by such procedural specification include the appointment of experts or arbitrators
to settle disputes, the processing of financial claims, the change of delivery times or quantities,
the early termination of the agreement and so on.
(iv)   Algebraic :  Such provisions specify formulae that are to be used to calculate values for
various parameters of the agreement, such as the price of goods during particular periods or
adjustments to prices or quantities of goods.
(v)  Effective:  Such provisions specify conditions under which other provisions apply.  These
are sometimes referred to as  secondary provisions.
The list is not intended to be an exhaustive or formal classification of contractual provisions.  A
particular contract clause may be of more than one of the aforementioned kinds.  For example an
individual provision may prescribe an obligation and at the same time define a term (and likewise for
the other kinds of functions).  Rather the purpose of the list is to indicate the kinds of clauses that
drafters put together when creating new contractual documents and to motivate the ensuing
discussion on what makes a provision ‘well-formed ’ .
Some criteria for well-formed provisions seem therefore to be:
(i)   Performance:  Is a given provision a good implementation of the corresponding part of the
agreement?  In other words, does it have the required effect or any undesirable side effects?
This consideration gives rise to the issue of  ambiguity : Where a given provision is formulated
in such a way that multiple interpretations can be ascribed to it, which is the one intended by
the parties?
(ii)  Consistency :  Is a given set of provisions concerning an aspect of the agreement consistent?
If more than one provision apply under some given circumstances, are they in conflict?
(iii)   Completeness:  Is a given set of provisions concerning an aspect of the agreement complete?
In other words, does it cater for all the intended circumstances that might arise?
(iv)   Redundancy :  Can one or more provisions be simplified to remove unwanted redundancy?
This is perhaps particularly the case for algebraic provisions where certain parameters in
formulae could be unfolded to yield a simplified formula.
Various techniques, practical and theoretical, have been developed that might be employed to address
such issues.  The following section reviews the most known ones and seeks to identify the extent to
which they can be usefully applied to legal contracts.
3   Techniques
Four approaches, which are to some extent interrelated in terms of their goals, are discussed.  Logic
programming representations offer executable specifications that are useful in addressing
performance, consistency and completeness.  Normalised Legal Drafting aims at removing unwanted
ambiguity.  Case Analysis provides a theoretical framework for analysing a set of norms with a view
5
to determine consistency, completeness and redundancy.  Finally, the Normative Positions approach
offers the means to address performance and consistency.  Of the four approaches, Logic
Programming is the one that does not purport to represent legal notions (such as duty, right,
obligation and so on) by employing any form of deontic logic.
3.1  Contractual Provisions as Logic Programs
A detailed representation of contractual provisions could follow in the same steps of a popular
approach for the representation of legislation or regulations where provisions are represented as logic
programs.  The most celebrated instance of this approach is the formalisation of the British
Nationality Act 1981 [Sergot et al. 1986].  Social security regulations [Hammond 1983] and Indian
civil service pension rules [Sergot 1991] are also amongst the areas where the approach was tried
successfully (for an overview of the approach cf. [Sergot 1988]).
3.1.1  Overview
The following example reproduced from [Kowalski 1995] illustrates the approach on a subsection of
the British Nationality Act 1981:
1(1) A person born in the United Kingdom after commencement shall be a British citizen if at the time of the
birth his father or mother is—
(a) a British citizen; or
(b) settled in the United Kingdom.
This can be paraphrased in logic programming form as follows, where predicates are in infix notation
and variables in upper case:
X acquires british citizenship by section 1.1
if  X is born in the uk at time T
and T is after commencement
and  Y is parent of X
and Y is a british citizen at T
or  Y is settled in the uk at T
The main feature of this approach is the close resemblance between the legislative or regulatory text
and its corresponding formulation as a logic program.  Hence, the representation of such text
maintains a clarity that renders its inspection, verification, modification and extension easier to
perform than on algorithmic representations.
Legislation or large sets of regulations typically contain heavy amounts of cross-references, with
special cases or exceptions to rules appearing at different parts of a document or even in different
documents; to attempt an algorithmic representation is a cumbersome task.  Moreover, legislative and
regulatory text may be ambiguous as it is expressed in natural language.  Such ambiguity may be
syntactic or semantic.  Sergot [1988] cites the example of regulations under which a woman could
receive a  ‘Housewives Non-Contributory Invalidity Pension’  (H NCIP), where one provision states
that a woman would be entitled to H NCIP “if she is incapable of performing her normal household
6
duties to a substantial extent”.  The scope of  ‘ to a substantial extent ’  is not clearly defined.
Moreover, the term itself is vague as what constitutes ‘ substantial extent ’ might be different things to
different people.  Open-textured concepts are abundant in legislative and regulatory text and an
advantage of the logic programming approach is that multiple interpretations may be supported
within the same representation as different logic programs.
The main benefit of logic programming representations of legislative or regulatory text is that they
constitute executable specifications.  Hence, drafters or policy-makers can ascertain whether the
provisions they are creating have the desired effect or any unwanted side effects by  ‘ executing ’  the
model.
3.1.2  Application
Having reviewed briefly the main features of the approach, let us now see to what extent it can be
usefully applied to the representation of contractual provisions.  As Sergot points out [1988],
provisions that are essentially descriptive in nature admit logic programming formulation in a
straightforward manner.  That is for provisions which specify conditions under which an entity X is
to be classified as being of type Y we can construct corresponding logic programs with relative ease.
Consider for example the following extract from a sample contract [IEE 1988]:
(46) Force Majeure means:-—war, hostilities (whether war be declared or not), invasion, act of foreign enemies;
—ionising radiations, or contamination by radio-activity from any nuclear fuel, or from nuclear waste from the
combustion of nuclear fuel, radio-active toxic explosive, or other hazardous properties of any explosive nuclear
assembly or nuclear component thereof;
—pressure waves caused by aircraft of other aerial devices travelling at sonic or supersonic speeds;
—rebellion, revolution, insurrection, military or usurped power or civil war;
—riot, civil commotion or disorder;
—any circumstances beyond the reasonable control of either of the parties.
It is not difficult to construct a logic program for classifying an incident as Force Majeure.  In fact,
such a representation may take various syntactic forms, one of which is shown below:
fm_circumstance(war).
fm_circumstance(hostility).
fm_circumstance(invasion).
:
force_majeure(Episode) 

fm_circumstance(Episode).
force_majeure(Episode) 

beyond_parties_reasonable_control(Episode).
Such a representation might be useful for drafters if they want to establish quickly what
circumstances are covered under such a definition—though with the last sub-clause any circumstance
7
is pretty much covered if it is deemed to be beyond the reasonable control of the parties, and what is
‘ reasonable control ’  is open-textured.
Logic programs can also be constructed for prescriptive provisions if we regard them as
qualifications, that is statements specifying legal relations for the parties in given conditions.  For
example, the following is a provision from the section on  ‘ Contractor’ s Obligations ’  of a sample
contract [IEE 1988]:
(14.4) The Contractor shall not without the Engineer’s consent make any material alteration to the approved
Programme.
This might be regarded as defining an obligation for the Contractor to refrain from performing a
certain action in given conditions.  A logic programming formulation of this clause might go along
the lines:
The contractor is obliged by section 14.4 to not alter the programme P
if  there does not exist consent by the engineer to alter the programme
P.
Consider also the following example from the same sample contract:
(14.1) Within the time stated in the Contract or, if no time is stated within 30 days after the Letter of
Acceptance, the Contractor shall submit to the Engineer for his approval the Programme for the execution of the
Works showing: -(a) the sequence and timing of the activities by which the Contractor proposes to carry out the Works
(including design, manufacture, delivery to site, erection and testing),
(b) the anticipated numbers of skilled and unskilled labour and supervisory staff required for the
various activities when the Contractor is working on Site,
(c) the respective times for submission by the Contractor of drawings and operating and maintenance
instructions for the approval thereof by the Engineer,
(d) the times by which the Contractor requires the Purchaser
(i) to furnish any drawings or information,
(ii) to provide access to Site,
(iii) to have completed any necessary civil engineering or building work (including foundations for the
Plant),
(iv) to have obtained any import licences, consents, wayleaves, and approvals necessary for the
purposes of the Works,
(v) to provide electricity, water, gas, and air on the Site or any equipment, materials or services which
are to be provided by the Purchaser.
This section specifies an obligation on the part of the Contractor to submit a Programme of a
specified description to the Engineer within a specified time.  A logic programming representation of
such an obligation might go along the lines:
The contractor is obliged by section (14.1) to submit an acceptable
programme P at time T
if T is the time specified in the contract.
The contractor is obliged by section (14.1) to submit an acceptable
programme P at time T
if there exist a letter of acceptance at time T1
and  T is later than T1 by 30 days.
8
The representation could be extended by an appropriate definition for what makes an acceptable
programme according to section (14.1).  This is certainly neither the only nor the best way to
represent the prescriptions above.  The purpose of the example is to illustrate a point made by Jones
and Sergot [1992]: Prescriptive statements can  be regarded as definitional and be represented as logic
programs if one wants to retrieve such information or to construct a system that operates as they
specify.  Where it is important to maintain the distinction between what ought to be the case and
what actually  is  the case, however, such representations fail, as they do not accommodate the
possibility that norms might in practice be violated.  The representation of section (14.1) above, for
instance, is useful if all we want to establish is what obligation the contractor has with respect to this
section, what action on his part is characterised by this obligation and when such obligation is to be
carried out.  There are no means of establishing what follows from such an obligation; for example
what happens if the contractor violates it and, relevant to that, what constitutes violation—the non-performance of the act within the specified time period, the non-conformance of the programme to
the standards set out in the contract or both?  Jones and Sergot conclude that in cases where we want
to reason with the distinction between the ideal and the actual some form of deontic logic is required.
3.2  Normalised Legal Drafting
Since 1957, Allen has been advocating the use of symbolic logic as a tool for drafting legislative text.
Allen’ s project is specifically targeted at ambiguity arises from natural language.  Allen [1982]
claims that a large amount of litigation based on written instruments can be traced to the drafter ’ s
failure to convey his meaning clearly.
Such ambiguity is not always of the same kind; sometimes it may be deliberate, because the drafter
wishes the legal text to be open to various interpretations for political or social reasons; yet,
frequently, such ambiguity is inadvertent and arises from the way natural language is used as a means
to formalise and represent the drafter ’ s intentions.  In addition, such ambiguity is not always the
result of what is written, but of what is omitted as well.  Allen classifies various kinds of imprecision
and concentrates on syntactic ambiguity that arises from the relations between sentences.  He
proposes a normalised language in which the different nuances of the meaning of logical connectives
and terms are reflected by the use of appropriately marked words.  The following table shows the
basis of his classification for the use of terms:
DEFINED TERM 1. INTENDED IN THE DEFINED SENSE
DEFINED TERM
o
2. Intended in the ordinary language sense
DEFINED TERM
a
3. Intended to be ambiguous about whether occurrence is in defined sense or
ordinary language sense
defined term 4.(a) Drafter thought about problem, intended to be ambiguous about whether
this occurrence is in defined sense or ordinary language sense, but does not
wish to tell audience about this deliberate choice
OR
(b) Drafter did not think about the problem of whether this occurrence is in
the defined sense or ordinary language sense.
Table 1 Allen’s [1982] classification of defined terms
9
As regards connectives used to express relations between sentences Allen [1957] proposes formal
definitions that can be used to disambiguate between, for instance, inclusive and exclusive
disjunction, or between the use of  “if …then ” statements and  “if and only if” statements.
In his framework, all natural language statements are normalised ( “systematically pulverized”).  The
process of normalisation assists both in the interpretation of legislative provisions and the drafting of
new ones.  In the case of the former, this is because implicit relations between propositions become
explicit by ascribing precise meaning to connectives and terms.  In the case of the latter, this is
because the drafter becomes aware of the different possible interpretations that might be ascribed to
his text and has the option to remove inadvertent ambiguity by trying to reformulate his text so that
the only possible interpretation is the intended one.
The following example illustrates the motivation and the result of their proposal.  It concerns the
University of Michigan lease termination:
The University may terminate this lease when the Lessee, having made application and executed this lease
in advance of enrolment, is not eligible to enrol or fails to enrol in the University or leaves the University at any time
prior to the expiration of this lease, or for violation of any provisions of this lease, or for violation of any University
regulation relative to Resident Halls or for health reasons, by providing the student with written notice of this
termination 30 days prior to the effective time of termination, unless life, limb, or property would be jeopardized, the
Lessee engages in the sales or purchase of controlled substances in violation of federal, state or local law, or the
Lessee is no longer enrolled as a student or the Lessee engages in the use or possession of firearms, explosives,
inflammable liquids, fireworks, or other dangerous weapons within the building or turns in a false alarm in which
cases a maximum of 24 hours notice would be sufficient.  [Emphasis added].
The clause consists of a single sentence with the ambiguous form:
A if A1 and A2 or A3 or A4 or A5 or A6 or A7 unless B1 or B2 or B3 or B4 or
B5 in which cases B.
Depending on how this is bracketed different interpretations arise.  Allen and Saxon identify
approximately 80 clarification questions that may be required in order to disambiguate between all
possible interpretations and as a result of such a process they conclude that the intended
interpretation is:
((A if ((A1 and (A2 or A3)) or A4 or A5 or A6 or A7))
if not (B1 or B2 or B3 or B4 or B5))
and (if (B1 or B2 or B3 or B4 or B5) then B).
and (if not (B1 or B2 or B3 or B4 or B5) then not B.
Here ‘when’  is taken to mean ‘ if ’ ,  ‘unless’  is taken to mean ‘ if not’ and  ‘ in which cases ’  is taken to
mean ‘ if and only if ’ .  Had both ‘when’  and ‘ in which cases’  be taken to mean  ‘ if ’  the resulting
interpretation would be:
((A if (A1 and (A2 or A3)) or A4 or A5 or A6 or A7))
if not (B1 or B2 or B3 or B4 or B5)
and if (B1 or B2 or B3 or B4 or B5) then B).
Kowalski [1995] argues that logic programming representations are superior because they render
many of the possible interpretations that might arise logically implausible.  Kowalski ’ s formulation
of the University of Michigan lease termination clause is:
A if A1 and A2 and not B.
A if A1 and A3 and not.
A if A4 and not B.
A if A5 and not B.
A if A6 and not B.
A if A7 and not B.
10
B if B1.
B if B2.
B if B3.
B if B4.
B if B5.
In this the condition not (B1 or B2 or B3 or B4 or B5) has been replaced by the simpler condition not
B (under the assumption that B1– B5 are the only conditions under which B holds).  Kowalski [1995]
points out however, that in logic programming there is no difference between the representation of
‘ if ’ and the representation of  ‘ if and only if ’ as the closed world assumption operates and the logical
consequences of a program include those of its completion.  Therefore, the two interpretations
produced by Allen and Saxon correspond to the same logic program.  If we were interested in
establishing the effects of a given set of provisions, the logic programming formulation above would
account for two of Allen ’ s interpretations.  Moreover, the logic programming representation can
easily be turned into a natural language expression that is much clearer than the original clause, and
more concise than the natural language expression corresponding to Allen and Saxon ’ s
representation.  Given the fact that logic programs provide executable specifications, as was stated in
the previous section, there does not seem to be any advantage to Normalised Legal Drafting over
Logic Programming.
More recently, [Allen & Saxon 1993] the framework was extended with explications of Hohfeld ’ s
[1913]  fundamental legal conceptions  to cater for normative statements.  These are the legal notions
identified by Hohfeld as the  “lowest common denominators” through which all other legal relations
may be defined.  Hohfeld organised them into pairs of jural opposites and jural correlatives and they
are summarised in the following tables; relations appearing in the same column are opposites
whereas relations along the same diagonal are correlatives.  Terms such as  ‘right ’ ,  ‘privilege ’ ,
‘ power’ ,  ‘ entitlement ’  and so on are often conflated whereas they entail different things as pointed
out by Hohfeld.
Right Privilege Power Immunity
No-right Duty Disability Liability
Table 2 Hohfeld ’ s Fundamental Legal Conceptions
Allen and Saxon’ s extended language for normalised legal drafting is the core of the MINT system in
which a lawyer-user is asked to determine his intended interpretation of a given statement when
presented by a set of alternatives produced by the system.  It is not altogether clear whether the user ’ s
choices at one point of the consultation session progressively narrow the space of available
interpretations, that is, whether the user ’ s commitment to some interpretation for part of the
legislative or regulatory text entails commitments to particular interpretations for other parts of the
text, by restricting choices to those that would be consistent with previous ones.  This is the kind of
functionality offered by Sergot ’ s [1998] automated analysis of normative positions.
11
3.3  Case Analysis
Carlos Alchourrón and Eugenio Bulygin [1971] provide an explication for consistency, completeness
and independence, the formal properties of normative systems.  Their framework is theoretical, in
that it is practically impossible to conduct extensive case analysis, but for very narrow domains.  It is
useful however, in that it provides a formal basis for characterising the consistency, completeness
and independence of a set of norms.
3.3.1  Elements of a normative problem
In Alchourr ó n and Bulygin ’ s framework, a normative problem is the deontic characterisation of a set
of actions  U A
  for an agent, that is the designation of individual actions in the set as permitted,
obligatory, forbidden and so on.  The set of situations in which such actions may take place is called
the  ‘universe of discourse ’   U D
: these are the situations for which the drafter or analyst of a set of
norms has to cater by specifying what actions are permitted, obligatory, forbidden and so on.  In
other words a normative problem requires the association of situations of U D
with deontically
characterised actions of  U A
; such deontically characterised actions are called  solutions .
The situations of the universe of discourse are essentially state-descriptions, that is, truth-functional
compounds of the properties that hold in them.  Alchourr ón and Bulygin’ s method requires the
identification of a finite set of properties  U P
  that make up situations.  Such properties are intended to
be the ones relevant to the normative problem and their selection plays a central role in the whole
approach as we shall see in the discussion section later.  Truth-functional compounds of properties of
the  U P
define the factual range of the problem, the set U C
  of elementary cases.  It can be shown that
every situation in the universe of discourse belongs to exactly one elementary case.  Put alternatively,
the universe of cases U C
  comprises schemata for cases and partitions the individual cases of the
universe of discourse in classes.  To provide solutions for the situations in  U D
  it suffices to provide
solutions for the elementary cases  U C
.
The set of solutions comprises deontically characterised truth-functional compounds of possible
actions from the U A
.  Possible characterisations are: Obligatory (O ), prohibited ( B), optional or
facultative ( F) and permitted ( P).  Deontic logicians have debated on whether these notions are inter-definable and on which should be taken as primitive.  Alchourró n and Bulygin take  P as the primitive
deontic operator and define the others as follows, where  α denotes an action:
•   O α  ≡
∆
  Pα  ∧   ¬P¬α
•   Bα  ≡
∆
  ¬Pα  ∧   P¬α
•   Fα  ≡
∆
  Pα  ∧   P¬α
These definitions capture some common-sense intuitions: What is obligatory is not prohibited and
vice versa; a facultative act is neither prohibited nor obligatory; an act is obligatory if its
complementary is prohibited.
12
The set of solutions U S
 to the normative problem comprises therefore expressions involving these
deontic operations.  The correlation of solutions with cases is done through the norms that are being
examined, that is the linguistic expressions of the legislative or regulatory text.  This is illustrated by
the following example taken from a sample contract.  The provisions concern relief from liability
under Force Majeure circumstances with unnecessary detail omitted for the sake of clarity and space:
(C1)   The expression “ Force Majeure”  shall mean any event or circumstance which is beyond the
control of the Party concerned, acting or having acted as a reasonable and prudent operator,
resulting in or causing the failure by such Party to perform any of its obligations under this
agreement, which failure could not have been prevented or overcome by the Party.
(C2)   The circumstances which shall be within the definition of Force Majeure (fulfilling the
requirements of clause C1) shall include but not be limited by the following: Acts of God, forces of
nature, epidemic and quarantine restrictions, land-sliding, lightning, earthquakes, fire, floods,
storms, tidal waves, strikes, lock-outs or other industrial disturbances, sabotage, acts of war
(etc.).
(C3)   Notwithstanding anything in clauses C1 and C2 the following events shall not be treated as Force
Majeure:
[NB:  List of situations that do not qualify as Force Majeure]
(C4)   A Party intending to seek relief under this article shall not be entitled to such relief unless such
Party shall
(a)   as soon as practicable, but within 7 days of the day upon which the Party first
knew of the failure to fulfil its obligation, notify the other Party and provide the
other Party with an interim report containing all relevant information
(b)   within 30 days provide the other Party with a full report which shall amplify the
information contained in the interim report and contain further information as the
other Party may reasonably require
(c)   upon request give or procure access for representatives of the other Party to
examine the scene of the event which gave rise to the failure and such access
shall be at the expense of the Party who failed
(C5)   Relief under this article shall cease to be available to a Party in respect of an event of Force
Majeure if it fails to take as soon as practical all necessary steps to rectify the cause of the failure.
For this example our Universe of Discourse  U D
 contains all situations in which any of the parties that
have come into the agreement fail to fulfil its obligations and seeks relief from liability attributing
such failure to Force Majeure.  The question we seek to answer in all those situations is whether or
not the party that failed to fulfil some obligation is granted relief from liability.
We must choose the properties that will be included in our Universe of Properties U P
   and to do so
we are guided by the clauses (C1)– (C5), which are relevant to the problem.  From clauses (C1) —
(C3) we see that the nature of the event, in which failure to perform an obligation occurs, is relevant,
as some circumstances are recognised as Force Majeure whereas others do not.  We can denote all
circumstances that qualify as Force Majeure ones by P.  From clause (C1) it also seems obvious that
the conduct of the Party seeking relief is also relevant, that is whether the Party acted as a reasonable
and prudent operator, and we can denote that by R.  The performance of certain actions by the party
seeking relief is relevant as can be seen by clause (C4) and we shall denote the property of those
actions being executed as Q.  Finally, rectifying the cause of a Force Majeure event is relevant to
whether relief is granted by clause (C5) and S denotes this.  The elementary cases to which these four
properties give rise are 16.
The action that is to be deontically characterised is providing relief from liability to a party due to
Force Majeure, here denoted by  ρ .
13
In contracts, there are at least two agents whose actions we want to characterise deontically.  Deontic
operators are therefore indexed by p (for the party granting relief), s (for the party seeking relief) and
x (where this distinction is not important).
The norms expressed by provisions (C1 – (C5) can be reformulated as follows, where each norm takes
the form <solution>/<case>:
(N1)   O p
ρ  / (P, Q, R, S)
(N2)   Bp
ρ  / (¬Q)
(N3)   Bp
ρ  / (¬S)
(N4)   Bp
ρ  / (¬P)
Their association with the elementary cases identified earlier is shown below:
C I
P Q R S N1 N2 N3 N4
C1
++++ O p
ρ
C2
+++– Bp
ρ
C3
+ + – +
C
4
++–– Bp
ρ
C
5
+ – ++ Bp
ρ
C
6
+ – + – Bp
ρ Bp
ρ
C
7
+ ––+ Bp
ρ
C
8
+ ––– Bp
ρ Bp
ρ
C
9
– +++ Bp
ρ
C
10
– ++– Bp
ρ Bp
ρ
C
11
– + – + Bp
ρ
C
12
– + –– Bp
ρ Bp
ρ
C
13
––++ Bp
ρ Bp
ρ
C
14
––+ – Bp
ρ Bp
ρ Bp
ρ
C
15
–––+ Bp
ρ Bp
ρ
C
16
–––– Bp
ρ Bp
ρ Bp
ρ
Table 3 Association of Elementary Cases with Solutions
Solutions that appear in the same column derive from the same norm that corresponds to that column.
Solutions that appear in the same row are the ones that can be derived from the system for the
elementary case that corresponds to that row.  A case which is not associated with any solution (i.e.
no solution appears in its row) is a normative gap (for example C
3
 above) .   As no solution is derived
for such an elementary case, no solution is provided by the system for any individual case of the
universe of discourse that belongs to the class of that elementary case.  A normative system with at
least one normative gap is called  incomplete.   A system is  inconsistent  in a case C
i
 if and only if there
are two or more incompatible solutions associated with it; otherwise, the system is consistent.  The
notion of incompatibility between two or more norms is relevant to the system of deontic logic that is
being used.  In the example given here, two norms would be incompatible if one were prescribing an
obligation for an agent to do  α and the other a prohibition for the same agent to do α under the same
circumstances.  A system is  redundant in a case C
i
if and only if the same solution is associated with
14
a case through two or more norms (i.e. it appears at least twice in the row associated with the case).
The system of the example is redundant as can be seen for example from C 6, C
8, C
10, C
12, C
13, C
14
,
C15 and C
16
.  The norms of a system are said to be  independent  if and only if there is no case in which
the system is redundant.  So the system of the example is incomplete, consistent and redundant and
only (N1) is independent of the other norms.  As Alchourr ón and Bulygin point out, though the
presence of redundant norms is undesirable, one must be careful when trying to adjust them lest
removal of a norm leaves normative gaps.
3.3.2  Discussion
Alchourrón and Bulygin ’ s approach offers an attractive theoretical framework within which to view
normative systems and their formal properties for legal drafting.  The criteria for well-formed
provisions can be addressed through an analysis such as the one conducted on the example.  Hence
the effects of a norm, whether there is ambiguity that needs clarification, the consistency and
completeness of a given set of norms can be addressed in the process of associating solutions with
cases.  It is not quite so easy to put their proposal to practice however and hopefully the example that
was presented has already raised some issues with the reader:
How are we certain that the properties chosen to define elementary cases are appropriate, or even the
only appropriate ones?  In the example, a simplification was made and various circumstances were
grouped under P or Q.  How can we be certain that we have chosen all of the relevant properties?
Alchourrón and Bulygin stipulate that the choice of properties is crucial and that they must fulfil
certain requirements:
They must be  logically independent  otherwise they will give rise to logically empty cases.  They
must also be empirically independent  otherwise they will give rise to empirically empty cases.  It is
possible for two properties to be logically but not empirically independent (that is, they may have a
causal relation).  A logically complete system is empirically complete as well so the requirement for
logical independence is stronger.  Properties must also be logically independent of the actions that we
are trying to characterise deontically, otherwise we might get solutions that would determine
deontically certain actions that are impossible to realise.  Finally the set of properties chosen and the
universe of discourse (the set of all situations that the system is meant to address) must be
corresponding, otherwise we can no longer rely on providing solutions for elementary cases in  U C
and completeness, consistency and independence of the norms for  U C
  do not necessarily entail
completeness, consistency and independence of norms for the universe of discourse.
A second point concerns the actual association of solutions with cases with reference to the linguistic
expressions on the legislative, regulatory, or contractual text.  The process itself relies on
interpretation.  For instance, in the example clause (C5) was taken to mean that a party is prohibited
from offering relief from liability if the other party does not rectify the cause of Force Majeure.  The
expression  “Relief under this article shall cease to be available to a Party…” however might be
regarded as implying that the party seeking relief has no right (in the Hohfeldian sense [Hohfeld
15
1913]) to expect to be granted relief, that is that there is no obligation on the other party to provide it.
The alternative interpretation of (C5) might therefore be  Fp
ρ  or  ¬O p
ρ .
A third point concerns the possible deontic characterisations for actions supported by the logic
employed by Alchourrón and Bulygin: Given an action  α the possible permission statements that can
be formulated for it and its complement are:
Pα  ∧   ¬Pα Contradictory
Pα  ∧   P¬α Fα
Pα  ∧   ¬P¬α O α
¬Pα  ∧  P¬α Bα or  O ¬α
¬ Pα  ∧  ¬P¬α Obligation to refrain from action
P¬α  ∧  ¬P¬α Contradictory
Contradictory statements aside, Alchourr ó n and Bulygin account for obligatory actions and their
dual, prohibited ones.  They also account for facultative actions but not for what might seem to be
their dual,  “obligations to remain passive ”.  What expressions of this kind essentially mean is that an
agent is not responsible for an action α or alternatively that  α is ultra vires for a given agent.  This
seems to deprive the framework from the ability to account for norms pertaining to agent
responsibility.  Recent systems of deontic logic do allow for this possibility as is the case in the
theory of normative positions, which is discussed later.
A fourth point concerns the action itself that is deontically characterised and the properties chosen
the representation as it stands does not account explicitly for temporal conditions.  Hence, what
might be intended by (N1) could be that it is obligatory to provide relief from liability after  a claim
has been made and during  the period in which Force Majeure circumstances obtain, rather than a
temporally unqualified obligation.  As regards temporally characterised norms themselves, we might
want to e,stablish the consistency and completeness of them with respect to time , which would call for
a different kind of analysis than the one presented here.  A lengthier discussion of these issues is
included in [Daskalopulu 1998] but the remarks made so far illustrate the complexity of the
approach.  As mentioned earlier it is practically impossible to conduct extensive case analysis but for
very narrow domains; even in those cases putting the approach to practice is far from straightforward.
3.4  Normative Positions
The theory of normative positions developed by Jones and Sergot [1993] and automated by Sergot
[1998] is an extended version of the Kanger – Lindahl theory [Kanger 1972; Lindahl 1977].  The
theory attempts to apply a combination of deontic logic and the logic of action/agency to the
formalisation of Hohfeld’ s [1913] fundamental legal conceptions.  The generalised theory included
automated inference methods which have been implemented in computer programs intended to
facilitate application of the theory to the analysis of practical problems, either for the purpose of
interpretation and disambiguation of legal texts, or in the design and specification of a new set of
norms.  The objective is to clarify and expand an incomplete and imprecise statement of requirements
16
into a precise formal specification at some desired level of detail.  The role of the system is to guide
this process, ensuring overall consistency and identifying any possibilities that remain to be explored.
3.4.1  Elements of the Theory of Normative Positions
Normative positions are formed by applying deontic operators to act-descriptions.  The relativized
monadic operator E x prefixes propositions with expressions of the form E
x
p read as  “agent x brings it
about that p ” or  “agent x sees to it that p is the case ”.  The operator is a success operator (that is, if x
brings it about that p, then p is the case) and is closed under logical equivalence.  Given a logical
statement describing some state of affairs  A, there are two possible fact positions:  A  and  ¬A.
From those three act-positions are constructed (cf. [Jones & Sergot 1993] for full details of the
method) for a given agent α:
(A1)   EαA
(A2)   Eα¬A
(A3)   ¬EαA  ∧   ¬Eα¬A
These are mutually exclusive and jointly exhaustive.  (A3) expresses the fact that an agent remains
passive or neutral with respect to A.  From those through a process of exhaustive generation and
elimination due to logical inconsistency, the method returns seven normative one-agent act positions
(which correspond to Lindahl’ s  “basic types of one-agent legal positions”).  P  and  O  are operators
for permission and obligation respectively and the axioms for them are those of SDL [Chellas 1980].
T1 PEαA  ∧   PEα¬A  ∧   P( ¬EαA  ∧   ¬Eα¬A)
T2 PEα
A  ∧   O ¬Eα¬A  ∧   P( ¬EαA  ∧   ¬Eα¬A)
T3 PEαA  ∧   PEα¬A  ∧   O (E
αA  ∨  E
α¬A)
T4 O ¬EαA  ∧   PEα¬A  ∧   P( ¬EαA  ∧   ¬Eα¬A)
T5 O EαA
T6 O ( ¬EαA  ∧   ¬Eα¬A)
T7 O Eα¬A
Table 4 Normative one-agent act-positions
These completely characterise an agent ’ s legal position with respect to a state of affairs.  For example
T1
 specifies that agent α is permitted to see to it that  A is the case, and he is permitted to see to it that
¬A is the case, and he is also permitted to do neither (in other words, the agent can do anything he
elects with respect to A and also he can choose to remain passive).  T
2
 specifies that the agent is
permitted to remain idle and to bring it about that A but is obliged to not bring it about that  ¬A.  By
T3,
 the agent has choice with respect to bringing it about that  A or that  ¬A but he must act.  T
6
 is
interesting with respect to a point made for Alchourr ó n and Bulygin ’ s analysis earlier: this requires
the agent to remain passive, which as we saw is not a possibility with the deontic logic employed in
Alchourrón and Bulygin ’ s framework.
17
Jones and Sergot have extended Lindahl ’ s theory to account for situations involving two agents (α
and  β), and this is particularly useful in contractual situations where some obligations or rights are
relevant to both parties rather than one of them only.  By re-naming T
1
– T7
 so that the agent they refer
to and the fact position they contain become explicit, two sets of one-agent act-positions are
obtained, T
1
[ α;A]– T7
[ α;A] and T
1
[ β ;A]– T7
[ β ;A].  There are 49 possible conjunctions describing the
legal positions of the two agents with respect to the state of affairs A, with one conjunct from the set
{T
1
[ α;A], …,T
7
[ α;A]} and the other from the set {T
1
[ β ;A], …,T
7
[ β ;A]}.  Of those only 35 (in Lindahl
denoted as R
i
) are internally consistent and constitute the space of possible legal relations that one
needs to consult in interpreting or disambiguating norms.
3.4.2  Discussion
To illustrate the normative positions approach to the analysis of norms, consider again an extract
from the sample contract concerned with Force Majeure, introduced in section 3.3.
(C4)   A Party intending to seek relief under this article shall not be entitled to such relief unless such
Party shall
(a)   as soon as practicable, but within 7 days of the day upon which the Party first
knew of the failure to fulfil its obligation, notify the other Party and provide the
other Party with an interim report containing all relevant information
(b)   within 30 days provide the other Party with a full report which shall amplify the
information contained in the interim report and contain further information as the
other Party may reasonably require
(c)   upon request give or procure access for representatives of the other Party to
examine the scene of the event which gave rise to the failure and such access
shall be at the expense of the Party who failed
(C5)   Relief under this article shall cease to be available to a Party in respect of an event of Force
Majeure if it fails to take as soon as practical all necessary steps to rectify the cause of the failure.
Let A  denote that relief from obligations under Force Majeure circumstances is granted,  B denote that
a party provides an interim report within 7 days,  C denote that a party provides a full report within 30
days, D  denote that a party gives access to the other party and  E denote that a party takes steps to
rectify the cause of Force Majeure.  Let the index  π  denote the party that provides relief from Force
Majeure and σ denote the party seeking relief from Force Majeure.  According to the clauses above
and the one-agent act-descriptions noted earlier, the normative content of these clauses is:
(N1)   ( O EσB ∧   OEσC  ∧   O EσD )  ∧  ( ¬ (E
σB  ∧  E
σC  ∧  E
σD )   ¬  O EπA)
(N2)   O EσE  ∧ ( ¬EσE    ¬  O EπA)
The expression  “A Party…shall not be entitled …” is captured by  ¬O EπA, conforming to the
Hohfeldian connection between right and duty (here right is taken to be entitlement and duty is taken
to be obligation).  The consequent of both norms could be different, depending on what the intended
interpretation is.  In the version given above, agent  π  is not obliged to offer relief from liability.
Other possible interpretations might be:
18
•   PEπA  ∧   PEπ¬A  ∧   O (E
πA  ∨  E
π¬A): agent  π  is permitted to see to it that relief is granted,
permitted to see to it that relief is not granted but obliged to see to it that relief is either granted
or not granted;
•   PEπA  ∧   PEπ¬A  ∧   P( ¬EπA  ∧   ¬Eπ¬A): agent π  is permitted to do nothing about the claim (which
seems unlikely as an intended interpretation but the text of the provision alone does not rule it
out)
•   O Eπ¬A: agent  π  is obliged to see to it that relief from liability is not granted.
There are two points worth discussing here.  First in the example it is not clear whether a party ceases
to be entitled to relief from liability by not complying with any or all of the three conditions set out in
clause (C4).  This is where Allen is right about ambiguity arising from the relations between
sentences.  If the conditions (a) – (c) are implicitly conjoined (which seems reasonable to assume) then
(N1) above is the intended interpretation.  Second, it might be argued that the party seeking relief is
not strictly speaking obliged to see to it that  A,  B,  C, and D  are the case.  That is, sub-clauses (a) – (c)
might be better conceived as procedural steps that a party seeking relief from liability due to Force
Majeure needs to complete in order for it to be granted such relief.  Such a reading of the clauses
would entail that a party performing these steps has the power to bring about a state of affairs in
which it is granted relief from liability due to Force Majeure.  Power here is meant in the Hohfeldian
sense, that is, it refers to the ability to create legal relations, in this instance the obligation on the
other party to grant relief.  Strictly speaking granting relief from liability itself is a power relation.
That is, the party granting relief from liability is establishing a state of affairs in which the relieved
party is not bound by its obligations.  If a party seeking relief from liability performs (a)– (c), then the
other party is obliged to exercise that power.
The main benefit of the theory of normative positions is therefore that it brings to the foreground
questions about intended interpretation, structural ambiguity, the precise nature of legal notions
(whether they are primary or concern other legal relations) and so on.  It is a more powerful tool in
addressing interpretation questions and in disambiguating between legal notions than any of the other
approaches discussed here and it can be integrated with the theoretical framework of case analysis to
address normative consistency, completeness and independence.  In a similar manner to Alchourr ón
and Bulygin ’ s framework, elementary cases are associated with normative positions for agents.  The
definitions for completeness and independence remain unaltered.  The definition for consistency
requires some adjustment:  A system is inconsistent  in a case C
i
 if and only if there are two or more
incompatible norms associated with it; otherwise, the system is  consistent.  For a given agent  α and a
fact position A, two norms are inconsistent if they are both members of the set {T
1
[ α;A], …,T
7
[ α;A]}.
For norms concerning two agents, two norms are inconsistent if they are both members of the set of
norms R
i
.  Where norms concern two agents but different fact positions, that is one is a member of
the set {T
1
[ α;A], …,T
7
[ α;A]} and the other a member of the set {T
1
[ β ;B], …,T
7
[ β ;B]} their
consistency depends on the relation, if any, between A  and  B.
19
Finally, the representation as it stands does not allow for temporally qualified normative statements.
These are very frequently encountered in contracts where parties have obligations that only arise
within specific times (whether these are in absolute or relevant terms).  Such an extension to the
representation would yield different definitions for consistency and completeness of norms.
4   Conclusions
This paper reviews four approaches that might be employed for the representation of contractual
contents at the micro-level, that is, at the level of detailed individual provisions.  The sample
contracts used for this research indicate that contractual provisions are of varying nature and serve
different purposes.  Some criteria for well-formed provisions are raised and they can be addressed in
different ways, depending on the nature of the provision to which they apply.  Logic Programming
representations offer the benefit of executable specifications that drafters may employ to determine
whether a set of provisions have the desired effect.  If we were aware of all possible circumstances
that might arise and executed them for the corresponding sets of data, we would be able to establish
completeness and similarly by querying the representation appropriately we could determine whether
there is inconsistency.  To determine completeness and consistency of temporal conditions of
provisions some sort of temporal algebra is required, such as Allen ’s [1984] for example.  Where
logic programming representations fall short is in accounting adequately for legal notions such as
duty, right, power and so on.  Though prescriptive provisions can be represented in the same way as
qualification norms for some purposes, such representation conflates the distinction between the
actual and the ideal.  Consequently, if the effects of prescriptive provisions are of interest, especially
in cases of violation, and if it is their consistency that one wants to establish, some form of deontic
logic is required.  The analysis of norms relative to the particular system of deontic logic adopted
merits a lengthier discussion than can be accommodated here.  Much of the literature in deontic logic
is concerned with the analysis of deontic conditionals and contrary-to-duty structures [Chisholm
1963], in which an agent ’ s obligation to bring about  A is conditional upon his violation of some other
obligation.  Such structures are very common in contracts, especially in relation to early termination
of an agreement.  As Jones and Sergot [1993] point out, even deciding what detachment principles to
accept for deontic conditionals raises non-trivial issues that need to be addressed.  There is a
significant amount of recent research in the representation of deontic conditionals and contrary-to-duty structures that needs to be put to the test in the domain of legal contracts (for example
[Alchourr ó n 1993; Jones & Pö rn 1991; Prakken & Sergot 1996; Prakken & Sergot 1997]).
Legal notions are explicitly defined in Allen and Saxon’ s normalised language for legal drafting, but
their system does not seem to offer inferencing facilities that would be useful in establishing the
effects of prescriptions.  Their proposed language is specifically aimed at identifying ambiguity,
especially the kind that arises by the use of connectives in natural language to relate segments of text
and by the conflation of certain legal notions, such as right, privilege, power.  The theory of
normative positions on the other hand can assist in disambiguating prescriptions, in establishing the
effect of given legal relations and in determining consistency of a set of norms.  Alchourr ó n and
20
Bulygin’s work provides a framework for the formal explication of norm consistency, completeness
and independence, though the underlying deontic logic is not as expressive as that of normative
positions.
Acknowledgements:  I am grateful to Marek Sergot of Imperial College, London for numerous inspiring
discussions over the last few years and for his endless patience and support.  A lot of the observations made in
this paper arise from Marek ’ s ideas though any omissions or possible misconceptions are entirely my own.
Thanks are also due to Chris Reed who read and commented on drafts of this paper, and to John Salt for
bringing the term ‘ ultra vires’  to my attention.
5   References
Alchourrón, C. E. (1993).  Philosophical Foundations of Deontic Logic and the Logic of Defeasible
Conditionals.  In Meyer & Wieringa (Eds.)  Deontic Logic in Computer Science: Normative System
Specification, John Wiley & Sons, Chichester.
Alchourrón, C. E. & E. Bulygin (1971).  Normative Systems.  Springer-Verlag, New York.
Allen, J. F. (1984).  Towards a General Theory of Action and Time.  Artificial Intelligence , 23.
Allen, L. E. (1957).  Symbolic Logic: A Razor-Edged Tool for Drafting and Interpreting Legal Documents .  The
Yale Law Journal, 66, 833 –879.
Allen, L. E. (1982).  Towards a Normalised Language to Clarify the Structure of Legal Discourse.  In Martino
(ed.) Deontic Logic, Computational Linguistics and Legal Information Systems, North-Holland.
Allen, L. E. & C. S. Saxon (1984).  Computer Aided Normalizing and Unpacking: Some Interesting Machine-Processable Transformation of Legal Rules.  In Walter (Ed.)  Computer Power and Legal Reasoning ,
West Publishing Company, 495–572.
Allen, L. E. & C. S. Saxon (1993).  A-Hohfeld:  A Language for Robust Structural Representation of
Knowledge in the Legal Domain to Build Interpretation-Assistance Expert Systems.  In Meyer &
Wieringa (Eds.)   Deontic Logic in Computer Science: Normative Systems Specification, John Wiley &
Sons, Chichester.
Atiyah, P.S. (1989).  An Introduction to the Law of Contract.   Clarendon Press (4
th
 edition), Oxford.
Bench-Capon, T. J. M. (1987).  Support for Policy Makers: Formulating Legislation with the Aid of Logical
Models.  In Proceedings of the First International Conference on Artificial Intelligence and Law ,
Boston Massachusetts, ACM Press, 181–189.
Chellas, B. F. (1980).  Modal Logic–An Introduction, Cambridge University Press, Cambridge.
Chisholm, R. M. (1963).  Contrary-to-duty Imperatives and Deontic Logic.  Analysis,  24, 33–36.
Daskalopulu, A. (1998).  Logic-based Tools for the Analysis and Representation of Legal Contracts.  Doctoral
Dissertation.  Imperial College London (forthcoming).
Daskalopulu, A. & M. J. Sergot (1995).  A Constraint-Driven System for Contract Assembly.  In Proceedings of
the 5th International Conference on Artificial Intelligence and Law , University of Maryland, College
Park, ACM Press, 62–69.
Daskalopulu, A. & M. J. Sergot (1997).  The Representation of Legal Contracts.  AI and Society , 11, 6 –17.
Fiedler, H. (1985).  Expert Systems as a Tool for Drafting Legal Decisions.  In Martino & Socci (Eds.)
Preproceedings of the Second International Conference on Logic, Informatics and Law , Florence:
IDG-CNR, 265–274.
Gordon, T. F. (1992).  A Theory Construction Approach to Legal Document Assembly.  In Martino (Ed.) Expert
Systems in Law, Elsevier Publishers.
Hammond, P. (1983).  Representation of DHSS  regulations as a Logic Program.  In  Proceedings   BCS   Expert
Systems ’ 83, Cambridge, British Computer Society.
Hohfeld, W. N. (1913).  Fundamental Legal Conceptions as Applied in Judicial Reasoning.  The Yale Law
Journal,  23, 16–59.
IEE (1988).  Model Form of General Conditions of Contract (MF1).  The Institution of Electrical Engineers,
London.
Jones, A.J. I. & Pörn, I. (1991).  On the Logic of Deontic Conditionals.  In Meyer & Wieringa (Eds.)
Proceedings of the First International Workshop on Deontic Logic in Computer Science, Amsterdam.
Jones, A. J. I. & Sergot, M. J. (1992).  Deontic Logic in the Representation of Law: Towards a Methodology.
Artificial Intelligence and Law, 1(1), 45 –64.
Jones, A. J. I. & M. J. Sergot (1993).  On the Characterisation of Law and Computer Systems: The Normative
Systems Perspective.  In Meyer & Wieringa (Eds.)   Deontic Logic in Computer Science: Normative
Systems Specification , John Wiley & Sons, Chichester.
Jones, A. J. I. & M. J. Sergot (1996).  A Formal Characterisation of Institutionalised Power.  Journal of the
Interest Group in Pure and applied Logics, 4:3, 429 –445.
Also in: Valdé s, Krawietz, von Wright, Zimmerling (Eds.),  Normative Systems in Legal and Moral Theory .
Duncker & Humblot, Berlin.
21
Kanger, S. (1972).  Law and Logic.   Theoria, 38.
Kowalski, R. A. (1995).  Legislation as Logic Programs.  In Bankowski et al. (Eds.)   Informatics and the
Foundations  of Legal Reasoning, Kluwer Academic Publishers, the Netherlands.
Lindahl, L. (1977).   Position and Change –A Study in Law and Logic.  Synthese Library,   112, D. Reidel,
Dordrecht.
Prakken, H. & Sergot, M.J. (1996).  Contrary-to-duty Obligations.  Studia Logica , 57 (1), 91 –115.
Prakken, H. & Sergot, M.J. (1997).  Dyadic Deontic Logic and Contrary-to-Duty Obligations.  In Nute (Ed.)
Defeasible Deontic Logic, Kluwer Academic Publishers, the Netherlands.
Sergot, M. J. (1988).  Representing Legislation as Logic Programs.  In Hayes, Michie & Richards (Eds.)
Machine Intelligence, 11, Clarendon Press, Oxford.
Sergot, M. J. (1998).  A Method for Automating the Analysis of Normative Positions.  Proceedings of the
Fourth International Workshop on Deontic Logic in Computer Science, Bologna, Italy.
Sergot, M. J., Kamble, A. S. & Bajaj, K. K. (1991).  Indian Central Civil Service Pension Rules:  A case study in
Logic Programming Applied to Regulations.  In Proceedings of the Third International Conference on
Artificial Intelligence and Law, Oxford, ACM Press, 118 –127.
Sergot, M. J., Sadri, F., Kowalski, R. A., Kriwaczek, F., Hammond, P. & Cory, H. T. (1986).  The British
Nationality Act as a Logic Program.   Communications of the ACM, 29:5, 370 –386.