Reasoning Between the Lines: A Logic of Relational Propositions


Dialogue & Discourse 9(2) 80-110   doi: 10.5087/dad.2018.203 

 

©2019 Andrew Potter 

This is an open-access article distributed under the terms of a Creative Commons Attribution License 

(http://creativecommons.org/licenses/by/3.0/). 

Reasoning Between the Lines: A Logic of Relational Propositions 

Andrew Potter APOTTER1@UNA.EDU 

Computer Science & Information Systems 

University of North Alabama 

Editor: Manfred Stede 

Submitted 08/2017; Accepted 12/2018; Published online 01/2019 

Abstract 
This paper describes how Rhetorical Structure Theory (RST) and relational propositions can be used 

to define a method for rendering and analyzing texts as expressions in propositional logic.  

Relational propositions, the implicit assertions that correspond to RST relations, are defined using 

standard logical operators and rules of inference.  The resulting logical forms are used to construct 

logical expressions that map to RST tree structures.  The resulting expressions show that inference 

is pervasive within coherent texts.  To support reasoning over these expressions, a set of rules for 

negation is defined.  The logical forms and their negation rules can be used to examine the flow of 

reasoning and the effects of incoherence.  Because there is a correspondence between logical 

coherence and the functional relationships of RST, an RST analysis that cannot pass the test of logic 

is indicative either of a problematic analysis or of an incoherent text.  The result is a method for 

analyzing the logic implicit within discursive reasoning. 

Keywords: Rhetorical Structure Theory, RST, relational propositions, propositional logic 

1 Introduction  

An important contribution of Rhetorical Structure Theory (RST) has been its usefulness in 

describing the organization of text.  An analysis created using RST presents the text structurally, 

identifying the functional relationships between the parts of the text and positioning them 

hierarchically to reflect their relative status within the overall textual organization.  This provides, 

among other things, a basis for examination of the nature of the coherence of the text, and it has 

led to diverse studies and applications in discourse analysis, argumentation, theoretical linguistics, 

psycholinguistics, and computational linguistics (Taboada & Mann, 2006a). 

An area of research that has received little attention has been the interrelationship between RST 

and propositional logic.  Although the relationship between logic and language has been of 

longstanding interest among philosophers, logicians, linguists, and psychologists, studies specific 

to RST and propositional logic are not to be found.  And yet the characteristics of RST and 

propositional logic suggest there may be untapped synergies.  For both, the object of analysis is a 

well-defined domain: in RST, the domain is the text under analysis (Mann & Thompson, 1987b), 

and in propositional logic, the domain is the universe of discourse, an explicit or implicit boundary 

within which the analysis is delimited (Boole, 1854).  Both involve analysis of relationships among 

discourse units at the sentential level, and both can be used to discover the sentential organization 

of discourse.  Both are concerned with discursive coherence, and both are useful in the analysis of 

discursive reasoning.   

The purpose of this paper is to examine the use of RST for rendering and analyzing texts as 

expressions in propositional logic.  The primary claim to be developed is that any text analyzable 



POTTER 

   81 

 

using RST can be restated as a Boolean expression.  That is, for any RST structure, there is a 

corresponding Boolean expression consisting of a set of variables representing the discourse 

elements and connected using logical operators.  The logical relations within such a Boolean 

expression are inferentially interdependent.  A set of rules of negation is formulated, and when 

applied to a text, they facilitate analysis of logical flows from one part to another.  This is useful in 

analyzing rhetorical structures and their logical consequences for the interpretation of text.  The 

result is a method for analyzing the logic implicit within discursive reasoning. 

RST is a functional theory of textual coherence.  It defines coherence in terms of the relations 

that occur between parts of the text.  The parts of the text may be elementary discourse units, or 

they may be structures of multiple related units. While there has been considerable debate as to the 

appropriate set of relations to be used in such an analysis (e.g., Carlson & Marcu, 2001; Grosz & 

Sidner, 1986; Hovy, 1990; Hovy & Maier, 1993; Knott, Oberlander, O'Donnell, & Mellish, 2001; 

Mann & Thompson, 1987b; Marcu, 2000; Sanders, Spooren, & Noordman, 1992; Stede, Taboada, 

& Das, 2017), there seems to be general agreement that there are identifiable relations and these 

may be used in describing the nature of coherence in a text.  RST further defines coherence using 

a set of constraints that define the structural characteristics of the text as a whole.  These constraints 

are identified as completeness, connectedness, uniqueness, and adjacency (Mann & Thompson, 

1987b).  The application of these constraints enables the analyst to identify the underlying 

hierarchical structure of the text.  

Logic defines coherence as consistency (Kolodny, 2007; Thagard, 1989).  To be considered 

coherent, the universe of discourse must be free of contradiction.  The universe of discourse 

specifies the assertions constituting the universe and their logical interdependencies.  The 

propositions constituting a coherent universe are true, either as elementary propositions asserted as 

axiomatic, or as inferred propositions, having been deduced either from elementary propositions or 

from other inferred propositions in accordance with the rules of inference.  Propositional logic 

includes a set of rules of inference, the logical functions that determine which conclusions may be 

drawn from a given set of premises (Boole, 1854; Russell, 1906).  Thus, a coherent universe of 

discourse may be viewed as an inferential structure of logically related propositions. 

These two concepts of coherence, one based on RST and the other on propositional logic, 

though distinct, have an important similarity: RST represents a text as discourse units connected 

by relations, and logic represents the universe of discourse as propositions connected by logical 

operators.  If it were the case that RST relations corresponded one to one with logical operators and 

units with propositions, the task of restating RST analyses as Boolean expressions would not be 

interesting.  However, the task is not so simple.  RST relations do not correspond one to one with 

logical operators.  As will be developed in Section 3, teasing out the logical expressions 

corresponding to RST relations involves a close analysis of the RST definitions as well as 

examination of examples of RST analyses.  Some of the resulting expressions are complex.  Of 

additional concern is the ostensible assumption of parity between elementary discourse units and 

propositions.  In logic, a proposition is defined as being either true or false, but not both, a definition 

normally associated with declarative sentences. But the discourse units of RST need not be 

declarative.  They can be any sentence type whatsoever, and in some cases, the units are sentence 

fragments or dependent clauses.  While this issue may be addressed in part through the recognition 

that many non-declarative units may be restated as propositions, this does not fully speak to the 

issue.   

To fully address this issue, it is first necessary to consider propositional logic within the larger 

context of Boolean algebra.  Propositional logic and the Boolean algebra are interchangeable 

conceptualizations.  Essentially, propositional logic is an application of the Boolean algebra.  The 

logical operations of the Boolean algebra correspond to the theorems of propositional logic, and 

Boolean variables correspond to elementary propositions.  Boolean logic is more general and lends 



REASONING BETWEEN THE LINES 

82 

 

itself to a variety of applications, including the design of switching circuits, mathematics, set 

theory, digital logic, and database query languages.  Although when discussing propositional logic 

we tend to speak in terms of truth values and truth functions, these semantics are logically arbitrary, 

and we could just as well speak of + and -, 1 and 0, yes and no, satisfiability and unsatisfiability, 

or any other bivalent conceptualization, such as belief and non-belief, positive and negative regard, 

desire and indifference, interest and disinterest, understanding and misunderstanding, or ability and 

inability.  To the extent that the primitives of RST can be understood in terms of bivalent values, 

they are amenable to logical treatment.  Any discourse unit that can be either accepted or rejected 

can be treated as a proposition and subjected to the rules of inference. 

Furthermore, in this paper, discourse units are of less concern as isolated units, but rather are 

of interest primarily with respect to the role they perform within a relational structure.  This is 

perhaps most easily understood in terms of relational propositions, the alter ego of RST relations 

(Mann & Thompson, 1986a, 1986b, 2000a).  Relational propositions will be explained in detail 

shortly, but for now, suffice it to say that they are implicit assertions that arise between clauses 

within a text and are essential to the effective functioning of the text.  They are RST relations 

reformulated as propositions.  A relational proposition consists of a predicate and two variables.  

The predicate corresponds to an RST relation, and the variables map to the units or spans the 

relation references, one to the nucleus, the other to the satellite. The predicate defines the 

relationship between these variables.  These propositions provide the basis for the logical 

interpretation of RST analyses.  While it is convenient to discuss relational propositions in terms 

of RST relations, as propositions, these relations may also be treated as truth-functional assertions.  

Thus, the perspective taken in this paper is that units can be considered not only within a broader 

Boolean conceptualization, but that their interpretation is reliant on the role they play as arguments 

within their relational propositions.  The value of this approach should become apparent through 

the analysis of specific relations and structures in Sections 3-5. 

An additional potential concern is with text spans.  In RST, there are, in addition to units and 

relations, text spans that represent uninterrupted linear intervals of text, consisting of multiple units 

and relations.  The Mann and Thompson (2000b) analysis shown in Figure 1 contains several such 

spans, such as the span represented as 2-4, representing the relations occurring among units 2, 3, 

and 4.  Such multiunit text spans are not explicitly defined in RST, raising the possibility of 

difficulties in logical interpretation.  The ELABORATION relation shown here is between one unit 

and an EVIDENCE span of two others. Mann and Thompson (2000a) claimed the corresponding 

relational proposition would be a complex and framework-dependent assertion.  However, the only 

framework upon which the corresponding relational proposition is dependent is Rhetorical 

Structure Theory itself.  The use of multiunit text spans is a diagrammatic shorthand for the 

structures encapsulated below them, and from the logical perspective, they function much the same 

as parentheses in a symbolic expression.  This will be discussed in detail in Section 2.2. 

 The organization of the remainder of this paper is as follows.  Section 2 presents brief 

overviews of the theoretical concepts used in the paper.  These include RST, relational propositions, 

and propositional logic. These overviews are followed in Section 3 with an analysis of RST 

relations, providing a specification of the logical form of each.  Section 4 presents a set of rules for 

negating relational propositions.  These rules are based on potential points of failure within the 

logical forms.  In Section 5, the logical forms and the rules of negation are used to examine the 

logical coherence of two texts.  Finally, Section 6 provides a perspective on this research, locates 

it within the larger context of studies in the relationship between coherence relations and logic, 

summarizes the results, and suggests some areas for future application and research. 

 



POTTER 

   83 

 

 
Figure 1. RST analysis containing multiunit spans 

2 Theoretical Background 

This section provides brief overviews of the major theoretical frameworks that support this paper.  

These include Rhetorical Structure Theory (RST), relational propositions, and propositional logic. 

RST provides the foundation for describing the sentential organization of a text.  It also provides 

the framework for using relational propositions.  The theory of relational propositions is an 

alternative conceptualization of RST, such that the relationships between text spans are viewed as 

implicit assertions that occur between clauses in a text.  Propositional logic can then be used to 

define these assertions as logical expressions constructed of propositions and logical operators. 

2.1 Rhetorical Structure Theory  

Rhetorical Structure Theory (RST) is a descriptive theory of text organization.  It is a tool for 

describing and characterizing texts in terms of the relations that hold among the clauses comprising 

the text.  The elements of RST are relations, schemas, schema applications, and structures.  An 

RST relation consists of three parts: the satellite, the nucleus, and the relation.  The satellite and 

nucleus are text spans.  The distinction between satellite and nucleus arises as a result of the 

asymmetry of the relations.  Within a relation, the nucleus is more salient than the satellite.  

Relations are defined in terms of four fields: constraints on the nucleus, constraints on the satellite, 

constraints on the combination of nucleus and satellite, and the intended effect.  For example, in 

the definition of the EVIDENCE relation, the constraint on the nucleus is that the reader might not 

believe the information presented in the nucleus, the constraint on the satellite is that the reader 

must believe it or at least find it credible, and the constraint on the combination of the nucleus and 

satellite is that by comprehending the satellite, the reader will have an increased belief in the 

nucleus.  The effect of EVIDENCE is that the reader’s belief in the nucleus is increased.  All of these 

fields must be satisfied for an application of the EVIDENCE relation.   

Mann and Thompson (1987b) proposed that relations could be grouped into two general 

categories: presentational and subject matter.  The distinction is based on intended effect.  As 

defined by Mann and Thompson (p. 18), presentational relations are relations “whose intended 

effect is to increase some inclination in the reader, such as the desire to act or the degree of positive 

regard for, belief in, or acceptance of the nucleus.”  An instance of a presentational relation is, when 

successful, an act of persuasion.  Subject matter relations are relations whose intended effect is that 



REASONING BETWEEN THE LINES 

84 

 

the reader recognize the relation in question.  Unlike the presentational relations, they involve no 

act of persuasion.  The significance of this distinction for determining logical form will become 

clear in Section 3. 

Schemas are abstract patterns used to specify the relations that may hold among a small set of 

text spans. Mann and Thompson (1987b) defined five schema types: 1) a relation with a nucleus 

and satellite, 2) multiple relations with a single nucleus and a satellite for each relation, 3) 

contrastive multinuclear relations, 4) joint multinuclear relations, and 5) sequential multinuclear 

relations.  The first schema, a relation with a nucleus and a satellite, is employed for most RST 

relations and is the schema most often found in texts.  Schema applications define the conventions 

for schema instantiation.  By convention, the application of schemas places no constraint on the 

order of spans (i.e. the nucleus may precede or follow the satellite), multi-relational schemas do 

not restrict which relations are used in their application, and any relation may be repeated multiple 

times in the application of a schema.   

The RST structure of a text refers to its overall composition in terms of schema applications.  

For the text to be judged coherent, it must conform to four constraints: completeness, 

connectedness, uniqueness, and adjacency.  To be complete, the text must be analyzable as a unified 

structure.  To meet the connectedness constraint, all text spans must be either minimal units or a 

constituent within a schema application.  Uniqueness requires that each schema application consist 

of a unique set of text spans.  The adjacency constraint states that the text spans comprising any 

schema application constitute one contiguous text span.  A consequence of these constraints is that 

an analysis performed using RST results in a tree structure.  RST has been used for a variety of 

purposes.  It has been used as a way of describing the relations between clauses of a text, 

irrespective of explicit grammatical or lexical signals.  It has been used for analysis of a wide range 

of text types, such as expository prose, narrative discourse, and news broadcasts.  It has been used 

in studies in contrastive rhetoric.  Of importance in this paper, RST has been used as a framework 

for investigating relational propositions, which are the topic of the next section.  

2.2 Relational Propositions  

Relational propositions are implicit coherence-producing assertions that serve to bind together the 

explicit parts of a text and are essential to the effective functioning of the text. This example 

(updated for the modern reader) from Mann and Thompson (1986b) is illuminating: 

 

I love to collect classic automobiles. My favorite car is my 1899 Duryea. 

I love to collect classic automobiles. My favorite car is my 2008 Toyota Camry. 

  

Even without knowing much about classic cars, most readers would agree that the first text is 

coherent and the second is not.  In the first text, there is an implicit relational proposition that binds 

the two parts of the text together.  The relational proposition asserts that the second part of the text 

is an elaboration of the first.  The reader does not need to be explicitly informed of this relationship.  

The parts go together.  In the second text, this is not the case.  The second part of the text does not 

follow from the first.  So essential are relational propositions that their failure may destroy the 

coherence of a text.  These propositions are pervasive throughout the text, and form the basis for 

various kinds of inference.  While lying outside mainstream RST research (Taboada & Mann, 

2006b), the existence and nature of these propositions has been generally acknowledged (Mann & 

Thompson, 1986a, 2000a; Nicholas, 1994; Taboada & Mann, 2006b).   

For each relation in an RST analysis, there is a corresponding relational proposition.  RST and 

relational propositions provide parallel accounts of coherence.  While RST accounts for coherence 

relations among the spans within a text, they do not identify the implicit relational acts that account 

for how the text functions (Mann & Thompson, 1986b).  That is the unique contribution of 



POTTER 

   85 

 

relational propositions.  Because relational propositions may be derived from the definitions of 

RST relations, they need not be sought directly in the text itself (Mann & Thompson, 1986a, 2000a).  

This means that relational propositions may be used as adjunct to RST relations.   

A relational proposition consists of a predicate and a pair of arguments.  The predicate 

corresponds to the RST relation, and the arguments correspond to the satellite and nucleus.  As 

presented by Mann and Thompson (2000a), the ordering of arguments in a relational proposition 

need not reflect their original ordering in the text, nor their canonical orderings in RST.   The 

satellite precedes the nucleus, such that (s)predicate(n).  For example, the relational proposition for 

the Duryea example above would be  

 

(My favorite car is my 1899 Duryea) IS AN ELABORATION OF (I love to collect classic automobiles) 

 

These expressions can be made more concise by using a conventional predicate notation with 

the initial argument as the satellite and the second argument as the nucleus; for example the 

relational proposition corresponding to the ELABORATION relation would be elaboration(s, n). 

For relational propositions to serve as truth-functional representations of RST structures, they 

not only need to be capable of expressing elementary relational assertions, they must be applicable 

to complete RST structures.  This is an area that has not been previously addressed.  As noted 

above, Mann and Thompson (2000a) claimed that representing multiunit text spans would be a 

complex framework-dependent undertaking, and did not address it.  They confined their research 

to simple relational propositions.  However, RST structures contain multiunit text spans.  For 

example the structure shown above in Figure 1, from (Mann & Thompson, 2000b), contains several 

multiunit spans (1-6, 2-6, 2-4, 3-4, and 5-6).  As noted earlier, such text spans are a diagrammatic 

shorthand for the structures appearing below them.  They provide the same organizational function 

as parentheses in a symbolic notation.  3-4 is one group of units, 5-6 is another.  There is a nesting 

of relational propositions that corresponds to the structure of the RST tree.  The EVIDENCE relation 

between units 3 and 4 is the satellite of the ELABORATION relation.  The CONCESSION relation 

between units 5 and 6 is the satellite of the INTERPRETATION relation, of which the ELABORATION 

relation is the nucleus.  The resulting structure is the nucleus of the PREPARATION relation: 

 

 preparation(1,interpretation(concession(5,6), elaboration(evidence(4,3),2))) 

 

Thus nested relational propositions enable the identification, not only of individual RST relations 

as assertions, but the expression of an entire RST structure as a relational proposition.  The 

evaluation of such an expression begins with the innermost relational propositions, and works 

outward through the nested propositions.  This suggests that there is a logical dependence that flows 

from the leaf nodes to the root.  Such an evaluation requires that each relation have a logical 

specification.  Supplying these specifications is the topic of Section 3 of this paper. 

2.3 Propositional logic 

In propositional logic, propositions are treated as primitive units that may be joined using logical 

operators to form more complex logical expressions.  These logical operators are used in this paper:  

 

Negation   p 

Conjunction  p  q 
Disjunction  p  q 
Exclusive Disjunction  p  q 

Material Implication → p → q 
Material Equivalence  p  q 



REASONING BETWEEN THE LINES 

86 

 

   

When propositions are combined using these operators, the resulting expressions depend on the 

constituent propositions and the semantics of the operators for their truth value.  If p is true then   

 p is false.  The conjunction of two propositions (p  q) is true if both p and q are true.  Disjunction 

can be defined in terms of negation and conjunction, as  ( p   q)), i.e., it is not the case that 

both p and q are false; (p  q) is true only if either p or q or both are true.  Exclusive disjunction (p 

 q) is true if p is true or q is true but not both p and q are true.  Material implication states that (p 

→ q) is true if it is the case that  (p   q).  That is, it will never be the case that p is true when q 

is false.  Material equivalence (or biconditionality) is a relation of mutual implication.  If ((p → q) 

 (q→ p)), then (p  q). 

For logical definitions of RST relations, discourse units within a satellite-nucleus relation are 

identified using the symbols s and n.  For example, the definition of ANTITHESIS, discussed below 

in section 3.1.1, is  

 

((s  n)  s) → n  

 

For structures involving more than two variables, the traditional symbols from propositional logic 

are used.  Thus, a structure including three discourse units identifies the symbols p, q, and r, as in 

 

(((p → q)  p) → q)  (((r →  q)  r) → q) 
 

And when providing a logical interpretation of a specific RST analysis, the units will be identified 

numerically, reflecting the numbering of units in the analysis, such as 

 

(((1 → 2)  1) → 2)  (((3 →  2)  3) → 2) 

 
The application of logic to RST, as with any application of logic to the real world, involves the 

use of judgment under conditions of uncertainty.  Mann and Thompson (1987b) specify that when 

specifying an RST relation between text spans, the analyst must make a judgment with respect to 

each of these spans, and that owing to the nature of text analysis, these judgments must be made in 

terms of plausibility rather than certainty.  Additional uncertainty will be introduced with the logical 

interpretations of the relations.  The RST relations, although specific, are applicable over a range 

of situations, and some will conform more closely to the interpretations than others.  Subject to 

these limitations, the following interpretation of RST relations will identify plausible inferential 

characteristics of rhetorical relations and show how they can be modeled using elementary 

operations of formal logic. 

3 A Logical Interpretation of Rhetorical Relations 

For RST to be useful in rendering texts as expressions in propositional logic, it is necessary to 

establish that for each RST relation there is a corresponding logical form.  In some cases this 

amounts to a simple one-to-one correspondence between the relation and a standard rule of 

inference, and in others the interpretation results in more complex inferential expressions.  For each 

RST relation it is necessary to examine the nature of the asymmetry between the satellite and 

nucleus.  In some cases the relationship is clearly implicative.  In others, discovery of the inferential 

quality of the relation requires more probing.  This section presents a discussion of each of the RST 

relations in the extended Mann and Thompson relation set, as identified on the RST Website,1 

                                                      
1 http://www.sfu.ca/rst/index.html  



POTTER 

   87 

 

providing a logical interpretation and rationale for each.  The relations are grouped according to 

their traditional categories, starting with presentational relations, followed by subject matter 

relations and then multinuclear relations.  

3.1 Presentational Relations 

Presentational relations are relations “whose intended effect is to increase some inclination in the 

reader, such as the desire to act or the degree of positive regard for, belief in, or acceptance of the 

nucleus” (Mann & Thompson, 1987b, p. 18).  An instance of a presentational relation is, when 

successful, an act of persuasion.  In each of the presentational relations, the satellite serves as the 

impetus for the reader’s acceptance of the intended effect.  In this way the writer seeks to influence 

the reader’s reasoning processes.  Increases in belief, desire, acceptance, regard, understanding, 

ability, and interest are the results of inferences specified by the relational propositions associated 

with these relations.  Consequently, these relations are logically argumentative; that is, the logical 

form of each relation presents a valid argument, consisting of a set of premises and a conclusion 

which follows from the premises.  A consequence of this is that the logical forms of the 

presentational relations are tautologies.  That these relations are tautologous arises due to the 

relationship between valid arguments and tautologies.  For any valid argument, there is a 

corresponding conditional statement whose antecedent is the conjunction of the premises and 

whose consequent is the conclusion, and this corresponding statement will be a tautology (Copi, 

1967).  Thus any relation that is logically argumentative will be tautologous.  The implications of 

this will be explored in detail in Sections 4 and 5. 

3.1.1 ANTITHESIS 

With the ANTITHESIS relation, the intended effect is to increase the reader’s positive regard for the 

situation presented in the nucleus.  The satellite is incompatible with the nucleus, such that the 

reader cannot have positive regard for both the nucleus and the satellite. This incompatibility 

increases the reader’s positive regard for the nucleus.  In this example of ANTITHESIS, from 

Thompson and Mann (1987), 

 

S: Rather than winning them with our arms, 

N: we'd win them by our example, and their desire to follow it. 

 

two alternatives are presented, with the nucleus clearly preferred over the satellite. The logic is 

simple, based on disjunctive syllogism: (((s  n)   s) → n).  To paraphrase, s or n but not s, 

therefore n.  

3.1.2 CONCESSION 

With the CONCESSION relation, the writer concedes the situation presented in the satellite and 

asserts that, though there might seem to be a potential incompatibility between the satellite and the 

nucleus, the satellite and nucleus are indeed compatible.  The writer holds the nucleus in positive 

regard, and by indicating a lack of incompatibility with the satellite, the writer seeks to increase the 

reader’s positive regard for the nucleus.  To understand the logic of CONCESSION, it is necessary to 

consider the situation in which it is presented.  In their paper on concessive relations, Thompson 

and Mann (1986, p. 441) observed: 

 

Only in terms of its discourse context can we understand how concession is a 'conceding' of 

something: it concedes the potential incompatibility of two situations in order to forestall an 

objection that could interfere with the reader's belief of the point the writer wants to make. 

 



REASONING BETWEEN THE LINES 

88 

 

The objection having been forestalled, its potential interference with the reader’s belief has been 

removed.  An example of CONCESSION occurs in the Common Cause letter, as analyzed by Mann 

and Thompson in several of their papers (Mann & Thompson, 1986a, 1986b; Thompson & Mann, 

1987): 

 

S: Tempting as it may be, 

N: we shouldn’t embrace every popular issue that comes along. 

 

The satellite and nucleus are not incompatible: one may be tempted, yet still resist embracing every 

popular issue.  As such, it is not the case that the satellite provides grounds for rejecting the nucleus: 

 

  (s →  n) 

 

And upon neutralizing this objection, the writer further invites the reader to infer from this the claim 

presented by the nucleus.  If the satellite does not imply the negation of the nucleus, then the nucleus 

holds.  The reasoning thus becomes an instance of modus ponens: 

 

 (( (s → n) → n)   (s →  n)) → n 

 

ANTITHESIS and CONCESSION both increase positive regard for the nucleus at the expense of 

the satellite.  ANTITHESIS negates the satellite, and CONCESSION dismisses it.  In analyzing texts, it 

can sometimes be difficult to determine which relation is applicable.  In the analysis shown above 

in Figure 1, Mann and Thompson coded the relation for text span 5-6 as concessive: although there 

is no recession, the economy is not doing all that well.  The antithetical reading is that the writer 

presents a choice between outright recession and protracted sluggishness, and ruling out recession 

leads to an inference of protracted sluggishness. The writer is negating the satellite while 

confirming the nucleus.  That fits the logical form of ANTITHESIS precisely.  In this manner, the 

logical forms can be useful in analyzing rhetorical structures.  This will be developed further in 

Section 5. 

3.1.3 EVIDENCE 

With the EVIDENCE relation, the satellite provides evidence in support of the nucleus. For the 

relation to achieve its intended effect, it is necessary that the reader accept the satellite, either 

axiomatically as an assumption, or inferentially derived from subordinate relations in the text 

structure.  The reader must recognize the implicative relationship between the satellite and the 

nucleus, with the result that the reader’s belief in the nucleus is increased.  In argumentative terms, 

the satellite is the ground and the nucleus is the claim.  The implicative relationship is subject to 

two constraints.  EVIDENCE must be defined to specify that if the antecedent is believable, the 

consequent will also be believable (material implication) and that the antecedent is believable 

(modus ponens).  Material implication specifies a conditional relationship between the antecedent 

and the consequent, such that if the antecedent is believable, the consequent will also be believable: 

p → q.  By definition, material implication does not specify whether the antecedent is true or not.  

As such, material implication is hypothetical.  The EVIDENCE relation is not hypothetical.  To 

achieve its effect, EVIDENCE requires that the antecedent (i.e. the satellite) be asserted.  

Consequently the implicative relationship is, in addition to material implication, modus ponens: 

 

 ((s → n)  s) → n 

 



POTTER 

   89 

 

Although the writer anticipates that the reader will accept the satellite, no such assumption is made 

about the nucleus.  If such an assumption could be made, there would be no need for the EVIDENCE 

relation: the relation could be more properly coded as ELABORATION.  In this example of the 

EVIDENCE relation, 

 

N: The program as published for calendar year 1980 really works. 

S: In only a few minutes, I entered all the figures from my 1980 tax return and got a result 

which agreed with my hand calculations to the penny. 

 

the writer is praising an income tax preparation software application; efficiency and consistency 

between programmatic and hand calculations are presented as evidence that the tax program really 

works (Mann & Thompson, 1987b).  The satellite is asserted and is in an implicative relationship 

with the nucleus.  As discussed below, this use of the modus ponens is applicable to the remaining 

presentational relations, including JUSTIFY, ENABLEMENT, MOTIVATION, BACKGROUND, 

PREPARATION, SUMMARY, and RESTATEMENT. 

3.1.4 JUSTIFY 

The satellite of JUSTIFY is used to increase the reader’s readiness to accept the writer’s right to 

present what follows in the nucleus.  This right to present may come from a variety of sources for 

use in a variety of situations.  For example, the writer may wish to assert a position of authority, 

make known educational qualifications or professional credentials, announce a leadership role 

within an organization, or cite qualifying experience within the area under discussion.  In this 

example from AI Magazine, Ken Forbus (2016) makes reference to his experience and expertise in 

artificial intelligence: 

 

S: Indeed, in my experience,  

N: today’s general-purpose AI systems tend to skate a very narrow line between catatonia 

and attention deficit disorder. 

 

His expertise is invoked to provide authority to the claim.  Without this authority, the comparison 

of AI systems to catatonia and attention deficit disorder would seem unworthy of attention.  Writers 

may also use JUSTIFY when seeking to set aside authority and speak inclusively, so as to obtain 

acceptance of what follows.  A good example of this is the famous quotation from US President 

John Kennedy, in which he used JUSTIFY to support an ANTITHESIS:   

 

S: And so, my fellow Americans: 

N: ask not what your country can do for you; ask what you can do for your country.  

 

Assuming an authoritative stance when asking the audience to give up their selfish desires in favor 

of altruistic contributions would be to risk accusation of self-serving hypocrisy.  JUSTIFY is also 

used when the writer is in a subordinate position, seeking permission to speak, or as a (perhaps 

ritualistic) show of respect, as when a lawyer in a U.S. courtroom begins an oral argument with the 

phrase, May it please the Court….  Stede et al. (2017) observe that with JUSTIFY the satellite may 

be used to present a basic attitude of the writer, making it easier for the reader to accept the nucleus.  

An example of this can be found in the Common Cause analysis, as shown in Figure 2.  The writer 

seeks to dispel any perception of conflict between his personal views and his strategic view 

regarding the nuclear freeze proposal.  JUSTIFY achieves credibility through advantageously 

positioning the writer with respect to the claim made in the nucleus. 

 



REASONING BETWEEN THE LINES 

90 

 

 

Figure 2. Common Cause example of JUSTIFY 

3.1.5 MOTIVATION and ENABLEMENT 

Mann and Thompson (1987b) treat MOTIVATION and ENABLEMENT as a subgroup within the 

presentational relations.  In both of these relations the nucleus presents an action.  The action may 

be presented as an imperative or as an implicit call to action.  MOTIVATION increases the reader's 

desire to perform the action, and ENABLEMENT informs the reader how to do it.  These two relations 

are often found together, their satellites linked to a common nucleus.  This structural arrangement 

has sometimes been referred to as a MOTIVATION-ENABLEMENT schema (Mann & Thompson, 

1987b).  The relations work well together, where one increases a desire and the other shows how 

to satisfy it.   

MOTIVATION and ENABLEMENT are implicative and follow the same logical form as 

EVIDENCE.  If the MOTIVATION satellite fails to motivate or if the ENABLEMENT satellite fails to 

increase the reader’s ability to perform the action, the action presented in the nucleus is less likely 

to occur.  When multiple satellites of a presentational relation share the same nucleus, they combine 

conjunctively to support the nucleus.  In the MOTIVATION-ENABLEMENT schema, desire and ability 

join forces to persuade the reader to take the indicated action.  

 

 
Figure 3. Linked argument using MOTIVATION and ENABLEMENT 

 

In the analysis shown in Figure 3, from Mann and Thompson (1987a), the action to be taken is 

to attend the 1983 Annual Staff Breakfast.  Information about when and where the event will be 

held enables the reader to attend, and the opportunity to meet other staff members, staff assembly 



POTTER 

   91 

 

representatives, and the university president provides reason for attending.  The absence of either 

satellite would not necessarily lead to failure to take action.  But given the presence of both 

satellites, if either of the relational propositions were to fail, i.e. if the intended motivation were 

demotivating, or the enablement contained incorrect or irrelevant information, then the 

MOTIVATION-ENABLEMENT schema would fail.  Taken together they present a linked argument, 

sharing a common locus of effect:  

 

 (((2 → 1)  2) → 1)  (((3 → 1)  3) → 1) 

3.1.6 BACKGROUND 

BACKGROUND is used to assure comprehensibility of the text.  A nucleus that would otherwise be  

incomprehensible is rendered understandable by virtue of its satellite.  In this example from Mann 

and Thompson (1987a), the satellite provides BACKGROUND for the nucleus.  Without the satellite, 

the nucleus would be difficult and perhaps impossible to understand: 

 

S: I am having my car repaired in Santa Monica (1522 Lincoln Blvd.) this Thursday 19th. 

N: Would anyone be able to bring me to ISI from there in the morning or drop me back 

there by 5 pm please? 

 

BACKGROUND has the same logical form as EVIDENCE, though the implication is one of 

comprehension rather than belief or acceptance. 

3.1.7 PREPARATION 

 Stede et al. (2017) specified that PREPARATION be used when the satellite serves no stronger 

purpose than setting the topic for the nucleus.  But setting the topic can be decisive in determining 

whether the writer achieves the intended effect of the text.  PREPARATION may determine whether 

the reader bothers to read the text.  In the analysis shown in Figure 4, the title serves to announce 

and delimit the topic of the text.  The satellite precedes the nucleus and tends to make the reader 

more ready, interested or oriented for reading the nucleus.  Used in this way, PREPARATION is 

intended to increase an inclination in the reader, namely an inclination to read the article.  Thus, 

PREPARATION is presentational and subject to the same logical form as EVIDENCE.   

 

 

Figure 4. The PREPARATION relation 



REASONING BETWEEN THE LINES 

92 

 

3.1.8 RESTATEMENT and SUMMARY 

The status of RESTATEMENT and SUMMARY as presentational relations is less than obvious.  They 

were originally defined as subject matter relations (Mann & Thompson, 1987b, 1988).  

Subsequently, on the RST Website, they were re-categorized as presentational, but their definitions 

were not updated to reflect this change.  With the RESTATEMENT relation, the satellite restates the 

nucleus, and the satellite and nucleus are of comparable length.  The SUMMARY relation is similar.  

The satellite restates the nucleus, but unlike RESTATEMENT, the satellite is shorter than the nucleus.  

The defined effect of both of these relations is that the reader recognizes that the satellite restates 

the nucleus.  If this were all there was to say on the matter, it would be difficult to see the intended 

effect as presentational – there would be no identifiable inclination in the reader that these relations 

increase.  Stede et al. (2017) argued that these relations should be treated as textual, and as with 

PREPARATION, there is clearly a textual dimension to their use.  However, going beyond the 

definitions, examining examples indicates that they are presentational.  RESTATEMENT and 

SUMMARY are used to reinforce a point the writer seeks to make.  In the following example of 

RESTATEMENT, from Mann and Thompson (1987b), the writer seeks to emphasize the claim that a 

clean car makes a statement about the character of the owner, and implicitly, that an unkempt car 

does too: 

 

N: A WELL-GROOMED CAR REFLECTS ITS OWNER 

S: The car you drive says a lot about you. 

 

This is not simply a case of organizing the text.  The RESTATEMENT drives home the point.  

SUMMARY can be used in the same way.  By recapping a complex text with a short summary, the 

writer can reinforce the claim.  Used in this way, RESTATEMENT and SUMMARY support reader 

comprehension, belief, or action in a manner similar to BACKGROUND and PREPARATION, and thus 

they adopt the same logical form. 

3.2 Subject Matter Relations 

Subject matter relations are relations whose intended effect is that the reader will recognize the 

relation in question (Mann & Thompson, 1987b).  Unlike the presentational relations, they involve 

no act of persuasion. The locus of effect resides in both the nucleus and the satellite.  Because the 

acceptability of the nucleus and the satellite is not in question, the minimal logical form for a subject 

matter relation could be considered to be the conjunction of the satellite and nucleus, s  n.   But 

the asymmetry of the subject matter relations indicate logical information beyond conjunction.  The 

intended effect is a recognition of some feature of the organization of the subject matter (Mann & 

Matthiessen, 1990), and the logical forms corresponding to the subject matter relations should 

capture such features.  For example, in a causal relation, the dependency between cause and effect 

may be represented as an implicative relationship.  However, unlike the presentational relations, 

the point here is not that the implication is needed to establish the acceptability of the consequent.  

The antecedent and the consequent are both presumed to hold, and the relation indicates that they 

are in an implicative relationship.  Such a relation would take the form, s  n  (s → n).  From a 

purely logical perspective, the implicative portion of this form is redundant: s → n will follow, by 

definition, from s  n.  But the implicative characterizes the inferential processes implicit within 

the intended effect.  The relational proposition corresponding to the logical form of an implicative 

subject matter relation is true only when both s and n are true and the implication holds.  Denial of 

the implication would be an indication of a failure to achieve the writer’s intended effect.  In other 

words, the text would be logically incoherent.  This is explored in detail in Sections 4 and 5.  The 

following sections provide definitions of the logical forms for each of the subject matter relations. 



POTTER 

   93 

 

3.2.1 Causal Relations 

The term cause can be defined broadly as any situation or activity that brings about a change in 

some other situation or activity.  In that sense, a causal relation is a relation that holds between 

parts of text, in which one part presents a cause and another part the effect.  From this perspective, 

there are several subject matter relations that describe cause and effect relationships.  Mann and 

Thompson (1987b) originally identified the causal relations as consisting of VOLITIONAL CAUSE, 

NON-VOLITIONAL CAUSE, VOLITIONAL RESULT, NON-VOLITIONAL RESULT, and PURPOSE.  They 

used the concept of volitionality as a means for distinguishing causes that involve the action of an 

agent, typically but not necessarily an action of a person, from causes that involve consequentiality 

without a chosen outcome.  Some analysts have elected to disregard this distinction between 

volitionality and non-volitionality.  Nicholas (1994) argued that since volitionality modifies only 

the nucleus or the satellite, but not the relationship between the two, it is rhetorically irrelevant.  In 

their RST annotation guidelines, Stede et al. (2017) ignored volitionality altogether and subsumed 

the Mann and Thompson relations under the more general categories of CAUSE and RESULT.  

Because volitionality has no bearing on the logical analysis, I have followed their practice in this 

discussion of causal relations.   

Stede et al. (2017) also specified that the RST relations CONDITION, OTHERWISE, and UNLESS 

relations should also be treated as causal.  These relations, along with PURPOSE, apply to 

hypothetical situations.  With CONDITION, the satellite presents a hypothetical, future, or otherwise 

unrealized situation upon which the nucleus depends.  With OTHERWISE, realization of the nuclear 

situation prevents realization of the satellite situation.  With the UNLESS relation, realization of the 

nucleus occurs provided that realization of the satellite does not.  With PURPOSE, the nucleus is an 

activity that will lead to the situation identified in the satellite.  

Two additional relations that can be considered as causal are MEANS and UNCONDITIONAL. 

MEANS specifies that the satellite presents a method or instrument which tends to make realization 

of the activity presented in the nucleus more likely.  In this regard it is akin to ENABLEMENT, except 

the presented activity is not intended to increase an inclination in the reader.  The satellite presents 

a situation that brings about the activity presented in the nucleus.  The UNCONDITIONAL relation 

asserts that the nucleus has no dependency on the satellite.  It is a relation of negative causality – 

that is, UNCONDITIONAL negates any causal relation between the satellite and the nucleus.  The 

following provides logical interpretations for each of these causal relations. 

3.2.1.1 CAUSE and RESULT 

CAUSE specifies that the satellite caused the nucleus, and RESULT specifies that the nucleus caused 

the satellite.  They are both subject matter implicative.  Some activity or situation arises as a result 

of some other activity or situation.  Their relational propositions assert that there is a causal 

relationship between the satellite and nucleus.  Here is an example of CAUSE excerpted from the 

Salvage and Archaeology analysis (available on the RST Website): 

 

S: Since the objects in a [ship]wreck represent a single moment in time, 

N: they provide better chronological information than even the most carefully excavated 

terrestrial site. 

 

If the analyst had viewed this relation as argumentative, if there were some doubt that objects in a 

shipwreck provide better chronological information, then the relation would have been coded as 

EVIDENCE.  But in the analyst’s judgment, this is not the case.  The intended effect is that the reader 

recognizes that the satellite is the cause for the nucleus.  Implication is needed to establish not the 

truth of the consequent, but that the antecedent and the consequent are in an implicative 

relationship.  The logical form of the CAUSE relation documents this implicative relationship: 



REASONING BETWEEN THE LINES 

94 

 

 

 s  n  (s → n) 
 

The RESULT relation is similar, except now the nucleus is the cause: s  n  (n → s).  In this 

example from Taboada and Renkema (2008), the speaker’s difficulty in smiling at Jesus arises as 

a result of the demands he places on her.  Hence the satellite follows from the nucleus: 

 

S: … I find it hard sometimes to smile at Jesus. 

N: He can be very demanding. 

3.2.1.2 MEANS 

The MEANS relation specifies that the satellite presents a method or instrument which tends to make 

realization of the activity presented in the nucleus more likely.  In this analysis from Taboada 

(2006), MEANS represents features of a method reflected in the design of the human body: 

 

N: ... the visual system resolves confusion 

S: by applying some tricks that reflect a built-in knowledge of properties of the physical 

world. 

 

Thus, MEANS is a special case of CAUSE, and it has the same subject matter implicative form. 

3.2.1.3 CONDITION 

With CONDITION, realization of the satellite results in realization of the nucleus.  Because the 

realization of the nucleus depends upon realization of the satellite, without realization of the 

satellite there will be no realization of the nucleus.  In this example from Mann and Thompson 

(1987b), the satellite specifies the conditions under which employees are urged to complete new 

beneficiary designation forms: 

 

N: Employees are urged to complete new beneficiary designation forms for retirement or 

life insurance benefits 

S: whenever there is a change in marital or family status. 

 

A change in marital or family status is the condition under which employees are urged to complete 

new beneficiary designation forms. The reader recognizes that the realization of the nucleus 

depends on the realization of satellite.  As long as the satellite remains unrealized, so will the 

nucleus.  Thus the relation is one of material equivalence (s  n). 

3.2.1.4 PURPOSE 

PURPOSE is logically similar to CONDITION.  With the PURPOSE relation the situation presented in 

the satellite is to be realized through the activity presented in the nucleus.  The satellite is an as yet 
unrealized result of the nucleus.  In this example from Mann and Thompson (1987b),  

 

S: To see which Syncom diskette will replace the ones you're using now 

N: send for our free "Flexi-Finder" selection guide and the name of the supplier nearest 

you. 

 

The purpose of sending for the selection guide is to determine the appropriate Syncom diskette.  

Without the activity identified in the nucleus, the situation identified in the satellite is without 



POTTER 

   95 

 

purpose and will not be realized.  The logic of PURPOSE is not only n → s, but also  n →  s.  So 

the logic is material equivalence, n  s.  

3.2.1.5 UNCONDITIONAL 

UNCONDITIONAL has received little attention in the literature, and some studies suggest that it 

occurs infrequently (e.g. Iruskieta, 2014; Rubin & Vashchilko, 2012).  This may be due in part to 

difficulties in differentiating it from CONCESSION.  Like CONCESSION, although the satellite could 

conceivably affect the realization of the nucleus, the nucleus is not dependent on the satellite.  

Figure 5 shows an example from an article on pre-performance anxiety (Brooks, 2013).  The 

nucleus holds, regardless of the value of the satellite: (s   s) → n. 

 

 
Figure 5. UNCONDITIONAL relation 

3.2.2 OTHERWISE and UNLESS 

The OTHERWISE and UNLESS relations are similar to one another.  With OTHERWISE, realization of 

the nuclear situation prevents realization of the satellite situation.  In preventing realization of the 

satellite situation, the nucleus effectively negates it.  In this example from the RST Website, 

 

N: Project leaders should submit their entries for the revised brochure immediately. 

S: Otherwise the existing entry will be used. 

 

Only by submitting their new entries can project leaders prevent the use of existing entries, and 

failure to submit will affirm the satellite.  Consequently, OTHERWISE indicates an exclusive 

disjunction, i.e., (n  s), or ((n  s)   (n  s)).  Either n will hold, or s will hold, but not both.  

UNLESS is similar to OTHERWISE, except now the realization of the satellite would prevent 

realization of the nucleus, rather than the nucleus preventing the satellite. 

3.2.3 SOLUTIONHOOD 

With SOLUTIONHOOD, the satellite poses a problem, for which the nucleus presents a solution.  The 

satellite may be presented as a question, request, problem, or other expressed need.  The logic of 

SOLUTIONHOOD is not to be found in the potentially interrogative nature of the satellite.  The logic 

can be derived from its RST definition and corresponding relational proposition.  To explain this, 

it is useful to examine an analogy between interrogatives and propositional functions, proposed by 

Hintikka (2007).  A propositional function is an “expression containing an undetermined 

constituent,” as defined by Whitehead and Russell (1910, p. 92).  Because of this undetermined 

constituent (or free variable), a propositional function is not itself a proposition, so it has no truth 

value and cannot serve as an element within an argument.  A propositional function defines the 

form of a proposition without asserting the proposition itself. The propositional function x 



REASONING BETWEEN THE LINES 

96 

 

specifies some predicate  of x without indicating whether there might be any x.  In this sense, 

according to Hintikka, a propositional function is a specification of a query.  But on the basis of 

such a query alone, it cannot be known whether any result will be forthcoming.  The query could 

return a null set.  The query can be made more specific by binding the variable x.  This binding, or 

existential quantification, is expressed as (x)x, which asserts that for at least one instance of x, 

x is the case, although the identity of x is unspecified.  As such it is a proposition.  But to specify 

that there is an identifiable x for x, the variable x must be replaced with a constant, v.  Since we 

know that (x)x, we can infer that there is some constant that will support this.  This is 

accomplished using existential instantiation: 

 

(x)x  

 v 

 

From the query (x)x, the answer, v is inferred.  With SOLUTIONHOOD, the writer has committed 

to providing a solution.  The problem is not simply posed and left for the reader to resolve.  The 

problem presented by the satellite is an indispensable part of a relation that must be satisfied by the 

solution presented as the nucleus.  The presentment of the problem, whether interrogative or 

otherwise, obliges the writer to provide an answer.  Otherwise there is no SOLUTIONHOOD.  The 

questions, requests, or problems posed in a SOLUTIONHOOD relation are not to be taken out of 

context.  The writer is providing a problem-solution matching set. 

The satellite is the specification of a query.  Since the writer is committed to providing a 

solution, the satellite is existentially quantified. The proposition (x)x represents the satellite of 

the SOLUTIONHOOD relation, where  is the problem, and x is the undetermined solution.  That is, 

it is asserted that there is a solution to the problem, but as yet the solution has not been identified.  

The solution is identified in the nucleus, where x becomes v, thanks to existential instantiation.  

Therefore, SOLUTIONHOOD is an inferential relation between the satellite and the nucleus.  Since 

(x)x and v are propositions, the inference may be restated propositionally as  

 

s  (s → n) 

 

Given the assertion of s, n.  This is not to claim that the answer to a question can be deduced from 

the question.  Rather that, in a coherent text, the posing of a problem can be understood to anticipate 

the solution, much as a database query uses a subset of first order logic to implicate the result set 

(Sowa, 2000).  An example will make this more concrete.  This example is from Mann and 

Thompson (1985): 

 

S: I'm hungry.  

N: Let's go to the Fuji Gardens. 

 

The problem is that the speaker is hungry, and the solution is that going to Fuji Gardens 

(presumably a restaurant) solves the hunger problem.  This relation makes some presuppositions 

about hunger and the means by which the discomfort that attends it may be relieved.  It is the desire 

to relieve the discomfort associated with hunger that gives the first statement its status as a problem, 

and it is the relief from discomfort resulting from the act of eating that gives the second statement 

its status as the solution to the problem.  It is through recognition of these presuppositions that the 

statements achieve coherence, and it is on this level that the inferential nature of the 

SOLUTIONHOOD relational proposition becomes apparent.  Expressed as a propositional function, 

the problem to be solved is some as yet unidentified source of relief (x).  Posed as the satellite to a 



POTTER 

   97 

 

SOLUTIONHOOD relation, the problem is existentially quantified, expressed as (x)relief(x).  The 

nucleus, using existential instantiation then substitutes Fuji Gardens (v) as the source of relief: 

 

(x)relief(x) → relief(FujiGardens)  

 

This is not the only effort to define SOLUTIONHOOD in terms amenable to inferential 

interpretation.  Abelen, Redeker, and Thompson (1993) treat Solutionhood as an interpersonal 

relation, which they define as a relation used for persuasion, and they categorized it alongside 

presentational relations such as EVIDENCE, JUSTIFY, and MOTIVATION.  Similarly, Azar (1999) 

cited the persuasive power of SOLUTIONHOOD in his study of the rhetorical structure of 

argumentative texts.  When leveraging the satellite to accept the nucleus, there would be an 

inferential dimension in SOLUTIONHOOD.  And while persuasiveness may not be a factor in the Fuji 

Gardens example, it can be found in other cases, such as the Syncom example used by Mann and 

Thompson (1987b), as well in Azar’s analysis.  But persuasiveness is not an essential feature of 

SOLUTIONHOOD, and for the purpose of discovering its underlying logic, it is unnecessary, as has 

been demonstrated here. 

3.2.4 ELABORATION 

With ELABORATION, the satellite presents additional detail about the situation or some element of 

subject matter which is presented or inferentially accessible in the nucleus.  Mann and Thompson 

(1987b) identified six subtypes of the relation, as listed in Table 1.  As noted by Taboada and Mann 

(2006b), any one of these subtypes is sufficient for the ELABORATION relation to hold, and in theory 

it could be possible to treat them as separate relations.  And while Carlson and Marcu (2001) did 

just that, for the purpose of logical interpretation this is unnecessary.  The common characteristic 

among the subtypes is that, as defined, the satellite presents additional detail about the nucleus.   

 

 

Nucleus Satellite 

Set Member  

Abstraction Instance  

Whole Part  

Process Step  

Object Attribute  

Generalization Specific  

Table 1. ELABORATION subtypes 

Whether through set membership, instantiation, mereology, process decomposition, object 

attribution, or specification, the satellite is logically implicit in the nucleus.  Were this subsumptive 

relationship inaccessible to the reader, the relevance of the satellite to the nucleus would be lost: 

the inference is necessary for the reader to recognize that the situation presented in the satellite 

provides additional information for the nucleus. In the example shown in Figure 6, in segment 1, 

the writer identifies an ongoing activity.  As analyzed by Mann and Thompson (1987a), segment 2 

elaborates this by identifying a type of activity (set-member), and unit 3 further details this activity 

(generalization-specific).  Each of the nuclei progressively circumscribe the satellites of their 

relations.  Thus ELABORATION presents a subject matter implication, with the nucleus as antecedent 

and the satellite as consequent: s  n  (n → s). 
 



REASONING BETWEEN THE LINES 

98 

 

 

Figure 6. ELABORATION nuclei circumscribe their respective satellites 

3.2.5 EVALUATION and INTERPRETATION 

EVALUATION and INTERPRETATION involve assessment of the situation presented in the nucleus.  

EVALUATION assesses the situation presented by the nucleus with respect to the writer’s positive 

regard.  INTERPRETATION relates the nuclear situation to a framework of ideas not involved in the 

nucleus and not concerned with the writer’s positive regard (Mann & Thompson, 1987b).  In both 

relations, the assessment presented in the satellite is a plausible inference made on the basis of the 

information provided in the nucleus.  To the extent that the inference offered by the satellite is 

plausible, the relation is implicative: ((n → s)  n) → s.  The implicative nature of these relations 

has been hinted at by several other researchers.  Both Stent (2000) and Stede (2008) note that 

EVALUATION involves the subjective judgment of the writer, and earlier Abelen et al. (1993) 

categorized both EVALUATION and INTERPRETATION as interpersonal relations.  Deriving their 

definition of the term interpersonal from Halliday and Hasan (1976), Abelen et al. view these 

relations as being notable for their persuasive power.  What distinguishes EVALUATION from the 

presentational relations is that, while the writer presents an expression of positive regard, the effect 

is not explicitly directed towards any prospective inclination in the reader.  The RST definition of 

the relation obscures this characteristic, because the effect occurs not in the nucleus but in the 

satellite.  In an excerpt from  Mann and Thompson’s (1987b) Syncom analysis, shown in Figure 7, 

the assessment presented in the satellite follows as a result of the information presented in the 

nucleus. An evaluation that did not follow from its nucleus would be incoherent.  Although 

INTERPRETATION does not deal in positive regard, the logical interpretation is identical to 

EVALUATION.  Unless the interpretation follows from its basis, it will be incoherent.  Just as with 

EVALUATION, the assessment presented in the satellite must be a plausible inference made on the 

basis of the information provided in the nucleus.   

3.2.6 CIRCUMSTANCE 

The satellite of the CIRCUMSTANCE relation sets the framework within which the reader is intended 

to interpret the nucleus.  In this respect, CIRCUMSTANCE is similar to BACKGROUND, as noted by 

Stede et al. (2017).  The framework imposed by the satellite functions as a constraint on the nucleus, 

delimiting it with respect to temporal, spatial, or other considerations. Figure 8 shows an analysis 

with two CIRCUMSTANCE relations, one spatial, the other temporal.  Sometimes an instance of 

CIRCUMSTANCE includes both spatial and temporal constraints, as in the example from the RST 

Website shown in Figure 9.  This serves to illustrate the constraining effect of the relation. 



POTTER 

   99 

 

 

 

Figure 7. The EVALUATION relation 

 

 

Figure 8. Two CIRCUMSTANCE relations, one spatial the other temporal 

 

Figure 9. One CIRCUMSTANCE relation with both temporal and spatial characteristics 

Although some researchers have found CIRCUMSTANCE to be non-causal (e.g., Carlson & 

Marcu, 2001), there are instances that combine temporal or spatial constraints with causality.  The 

following examples use causal CIRCUMSTANCE, the first from Mann and Thompson (1987b) and 

the second from Stede et al. (2017).  In the Visitors Fever example, the writer attributes the cause 

of the ailment to a visit with relatives in the Midwest.  In the second example, deregulation of the 

electricity market provides the reason for squeezing suppliers: 

 



REASONING BETWEEN THE LINES 

100 

 

N: Probably the most extreme case of Visitors Fever I have ever witnessed was a few 

summers ago 

S: when I visited relatives in the Midwest. 

 

S: When Veag came under pressure because of the deregulation of the electricity market, 

N: they compensated for this by squeezing their suppliers. 

 

While CIRCUMSTANCE is similar to ELABORATION with respect to the satellite providing 

additional information about the nucleus, the nature of the additional information is quite different.  

Rather than the nucleus subsuming the satellite, the effect of the additional information provided 

by the satellite circumscribes the scope of the nucleus.  The nucleus holds within the domain 

specified by the satellite.  The satellite specifies a condition under which the nucleus holds.  The 

logical form of CIRCUMSTANCE is subject matter implicative: s  n  (s → n). 

3.3 Multinuclear Relations 

Most of the multinuclear relations are logical conjunctions, with exceptions being DISJUNCTION 

and MULTINUCLEAR-RESTATEMENT.  DISJUNCTION, as the name suggests, is disjunctive.  The 

elements of MULTINUCLEAR-RESTATEMENT are materially equivalent assertions.  While the 

conjunctive nature of CONJUNCTION, CONTRAST, JOINT, and LIST should seem readily apparent, 

SEQUENCE might seem serially inferential.  With SEQUENCE, a succession relationship between the 

situations is presented.  Mann and Thompson suggest recipes as good examples.  There is a serial 

dependency among the nuclei.  The dependency could be temporal, process, or any other series.  

To the extent that this is the case, failure of any nucleus may render the successors unreachable.  

But this means that for a sequence to hold, all of its members must hold.  Hence SEQUENCE may 

be defined as the logical conjunction of its members. 

Mann and Thompson (1985) claimed that relational propositions are always present in multi-

sentence texts.  And yet none of their work on the topic mentions multinuclear relational 

propositions.  I propose that they would take the form of LIST(n1, n2, n3…), which would transform 

into the logical expression n1  n2  n3…  And DISJUNCTION would follow the same pattern, using 

the disjunction logical connector.  MULTINUCLEAR-RESTATEMENT seems to occur as pairs of 

biconditional nuclei, although in principle there could be larger sets. 

4 Negating Relational Propositions 

The logical forms defined in Section 3 can be used to construct logical interpretations of a text.  

This is accomplished by restating the RST structure of the text as a nested relational proposition 

and mapping the predicates comprising the proposition to their respective logical forms.  This 

suggests the possibility of analyzing texts as truth functional expressions.  If so, we would expect 

that these expressions would be true or false, and that the determinants of their truth value would 

be indicative of the logical coherence of the discourse.  This presupposes the ability to negate 

logical forms.  And yet, based on the definitions presented in Section 3, it would seem that all 

relational propositions must be true.  Presentational relational propositions are true because they 

are tautologies.  Subject matter relational propositions are true because their satellites and nuclei 

are non-controversial.  And while this observation—that relational propositions are by definition 

true—follows from the logical forms, and while these forms are in turn derived from their 

respective RST definitions, this clearly conflicts with both intuition and experience.  RST 

analyzability is no guarantee of truth or logical coherence.  As Mann and Thompson (1986b) and 

Marcu (1996) have shown, it is possible to assign plausible RST relations for incoherent discourse.  

The apparent difficulty arises as a result of a disparity between the writer’s intended effect and 

the actual effect on the reader.  The RST relations are defined to indicate the writer’s intended 



POTTER 

   101 

 

effect.  This is consistent with theories of rhetoric in general (Heuboeck, 2009), but when reduced 

to its logical essence manifests itself as self-affirming reasoning, since it assumes that the effect 

has been achieved.  Thus it is necessary to specify not only the forms of intended effect but their 

truth conditions as viewed from a skeptical perspective as well.  The perspective of a skeptical 

reader can be used to identify potential points of failure within the logical forms.  The potential 

points of failure can be used in constructing rules of negation for use in reasoning over logical 

forms.   

The fundamental differences between presentational and subject matter relations require 

separate approaches for accomplishing this.  The intended effect of presentational relations, to 

increase some inclination in the reader, is identified in the nucleus.  If the writer is successful, i.e., 

if the reader accepts the effect, the reader accepts the situation as presented in the nucleus.  The 

logical forms of the presentational relations build-in this acceptance.  This is evident in the 

implicative structure of these relations.  Presentational relational propositions are logically valid 

arguments.  The premises of the argument are located in the left hand side (LHS) of the logical 

form, to the left of the outermost implication.  From these premises, the right hand side (RHS) or 

conclusion, consisting of the nucleus, is deductively inferred.  Thus the tautological structure of 

these relations presumes success for the writer’s intended effect.  While presenting the logical forms 

from the writer’s perspective is consistent with the RST definitions, for the skeptical reader this 

amounts to begging the question.  For the skeptical reader, it is the soundness of the LHS that is of 

interest.  Although negation of the LHS will not necessarily negate the RHS, an LHS once negated 

provides no functional support for the intended effect.  Since the burden of persuasion is on the 

LHS, its negation is sufficient for rejection of the RHS.  Were this not the case, the relation would 

be subject matter rather than presentational.  The LHS of a presentational relation consists of a 

major premise and a minor premise, as identified in Table 2.  Denial of the LHS of the form can 

consist of the negation of either or both of the premises, or of the RHS itself.    

 

 

Relation 

LHS RHS 

Major Premise Minor Premise Conclusion 

ANTITHESIS (s  n)  s n 

CONCESSION ( (s →  n) → n)  (s →  n) n 

EVIDENCE, JUSTIFY, 

MOTIVATION, ENABLEMENT, 

BACKGROUND, PREPARATION, 

RESTATEMENT, SUMMARY 

(s → n) s n 

Table 2. Presentational relations summary 

The LHS of the ANTITHESIS relation is ((s  n)   s), which implies the RHS, n, such that ((s 

 n)   s) → n.  The LHS includes several options for disputation by the skeptical reader.  The 

negation of the LHS,   ((s  n)   s) follows whenever s holds (the minor premise is negated), n 

does not hold, or when the major premise is negated: 

 

(s   n   (s  n)) →  ((s  n)   s) 

 

For CONCESSION to succeed it is necessary that the satellite be compatible with the nucleus.  

The CONCESSION LHS is negated whenever the major or minor premise is negated, or when n does 

not hold.  Negating the minor premise challenges the compatibility between the satellite and the 

nucleus:   

 



REASONING BETWEEN THE LINES 

102 

 

( ( (s →  n) → n)  (s →  n)   n) →  (( (s →  n) → n)   (s → n)) 

 

Following the same approach, EVIDENCE is negated when the major or minor premise is 

negated, or when the nucleus is negated.  Negating the major premise denies the implicative 

relationship between the satellite and the nucleus.  Negating the minor premise (s) presents a valid 

argument for negation of the LHS.  These same rules of negation also apply to the JUSTIFY, 

MOTIVATION, ENABLEMENT, BACKGROUND, PREPARATION, RESTATEMENT, and SUMMARY 

relations: 

 

 (s → n)   s   n →  ((s → n)  s) 

 

As defined in Section 3, several of the subject matter relations are subject matter implicatives.  

For some of these, the implication flows from satellite to nucleus  (CAUSE, MEANS, 

CIRCUMSTANCE,  SOLUTIONHOOD), and for others the flow is from nucleus to satellite (RESULT, 

ELABORATION, EVALUATION, INTERPRETATION), but they all share the same general logical form, 

consisting of a conjunction of the satellite and nucleus and an implicative relation between the two.  

The negation rule for the s → n group is 

 

( s   n   (s → n)) →  (s  n  (s → n)) 
 

and for the n → s group, the rule is 

 

( s   n   (n → s)) →  (s  n  (n → s)) 
 

Since the logical forms of CONDITION and PURPOSE are material equivalence (s  n), negation 

requires that either the satellite be false while the nucleus is true, or the nucleus be false while the 

satellite is true: 

 

((s   n)  ( s  n)) →  (s  n) 

 

UNCONDITIONAL asserts that the nucleus is not dependent on the satellite, in the sense that n, 

regardless of the value of s, or (s   s) → n.  Negation of n suffices to negate the relation: 

 

  n →  ((s   s) → n) 

 

OTHERWISE and UNLESS, both being exclusive disjunctions, are negated when either both the 

satellite and the nucleus are true or when both are false:  

 

 ((n  s)   (n  s)) →  ((n  s)   (n  s)) 

5 The Logic of Relational Propositions 

The logical forms and rules of negation can be used in exploring the coherence of texts.  More 

precisely, these forms and rules can be used to explore textual reasoning as reflected in an RST 

analysis.  Any RST analysis can be restated as a nested relational proposition, and this nested 

relational proposition can be restated as a logical expression.  As a simple example, the Music Day 

analysis from Mann and Thompson (1987b), shown in Figure 10, can be restated as this relational 

proposition: justify(concession(2,3), 1).  The CONCESSION relation is the satellite of the JUSTIFY 

relation.  Because the logical form of JUSTIFY is modus ponens (((s → n)  s) → n), the nested 



POTTER 

   103 

 

concession(2,3) appears twice, first as the conditional part of the major premise and again as the 

minor premise, as shown in Figure 11. 

 

 

Figure 10. The Music Day analysis 

 

 
Figure 11. Music Day logic 

Since both CONCESSION and JUSTIFY are presentational relations, for the purpose of this analysis, 

only the LHS of the forms are of interest.  Here is the nested relational proposition reduced to the 

LHS of each relation: 

 

((((( (2 →  3) → 3)   (2 →  3)) → 3) → 1)  ((( (2 →  3) → 3)   (2 →  3)) → 3)) 
 

To this we can selectively apply rules of negation.  For example, if the reader challenges the 

CONCESSION relation by denying the compatibility between its nucleus and satellite, the result is 

negation of the entire expression: 

 

(2 →  3) →  ((((( (2 →  3) → 3)   (2 →  3)) → 3) → 1)  ((( (2 →  3) → 3)   

 (2 →  3)) → 3)) 

 

The negation is systemic because the negated CONCESSION is the minor premise of the JUSTIFY 

modus ponens argument.  Within an RST analysis, when a rule of negation can be plausibly applied 

to a relational proposition, this is an indication of a potential misalignment either between the logic 

and the rhetorical coherence of the text or between the logic and the RST analysis.  Thus the effects 

of faulty reasoning on the part of the writer and problematic RST encodings on the part of the 

analyst can be examined using the logic of relational propositions.  The following presents a logical 

analysis of a more complex example. 



REASONING BETWEEN THE LINES 

104 

 

The RST analysis to be considered is “Bouquets in a Basket,” shown in Figure 12.  According 

to Mann and Thompson (1987a), the text is expository.  The RST analysis consists primarily of 

subject matter relations.  The text begins with the statement, There is a gardening revolution going 

on (1).  Units 2-3 elaborate on the revolution mentioned in 1.  The structure consisting of 1-3 is the 

satellite of a BACKGROUND relation.  The nucleus of the BACKGROUND consists of a PURPOSE 

relation.   Mann and Thompson mention that unit 4, the satellite of a PURPOSE relation, presents a 

possible goal, of creating "your own" bouquet, with segments 5-8 providing the method. 

 

 
Figure 12. RST Analysis of the Bouquets in a Basket text 

The logic of the text, as implicit in this RST analysis, may be summarized as follows:  As the 

nucleus of an ELABORATION relation, the statement that there is a gardening revolution going on 

(1) is the antecedent of an implicative relation with units 2-3.  By definition of ELABORATION, that 

people are planting flower baskets with living plants, and mixing many types in one container for 

a full summer of floral beauty follows from the assertion that there is a gardening revolution going 
on: 

 (1  (2  3  (2 → 3))  (1 → (2  3  (2 → 3)))) 

 

By definition of BACKGROUND, from 1-3, it follows that the reader may create a bouquet of flowers 

(4) using the method presented in 5-8.  And by definition of PURPOSE, the method presented in text 

span 5-8 is materially equivalent with 4:  

 

 (((5  6  (5 → 6))  (5  (7  8  (7 → 8))  (5 → (7  8  (7 → 8)))))  4) 

 

Since BACKGROUND is a presentational relation, and it is an overarching relation for this text, the 

logical expression representing the text is this tautology: 

 

((((1  (2  3  (2 → 3))  (1 → (2  3  (2 → 3)))) → (((5  (7  8  (7 → 8))  (5 → 

(7  8  (7 → 8))))  (5  6  (5 → 6)))  4))  (1  (2  3  (2 → 3))  (1 → (2  3  

(2 → 3))))) → (((5  (7  8  (7 → 8))  (5 → (7  8  (7 → 8))))  (5  6  (5 → 6))) 

 4)) 

 
Thus, according to the RST analysis, the text is logically coherent.  This would be expected of any 

RST analysis, since the logical forms arise from the RST definitions.  The logic becomes more 



POTTER 

   105 

 

interesting when the rules of negation are applied.  There are a few inferences within this RST 

analysis that may seem problematic to the skeptical reader.  First, as a subject matter relation, 

ELABORATION assumes that the reader will accept both the satellite and nucleus.  If either is 

questionable, some other relation must be applied.  Taken on its own, there is a gardening 

revolution going on (1) will seem puzzling to readers if the hyperbolic use of the term revolution 

is not obvious.  In that case, substantiation would be needed.  If for lack of substantiation the 

statement is denied, the consequences are systemic, and the LHS of the BACKGROUND relation is 

negated, resulting in this tautology: 

 

(1 →  (((1  (2  3  (2 → 3))  (1 → (2  3  (2 → 3)))) → (((5  (7  8  (7 → 8)) 

 (5 → (7  8  (7 → 8))))  (5  6  (5 → 6)))  4))  (1  (2  3  (2 → 3))  (1 → (2 

 3  (2 → 3)))))) 

 

However, the lack of substantiation is not within the text, but emerges only when accessing the 

logic of the text through the intermediation of the RST analysis.  That is, the logic of the 

ELABORATION relation is inconsistent with the logical coherence of the text.  The satellite, rather 

than following from the nucleus, is needed to give credence to the claim asserted by the nucleus.  

This can be satisfied using EVIDENCE rather than ELABORATION for the relation between 1 and 2.   

A second problematic inference is found in the PURPOSE relation.  It is significant that the 

satellite of this relation, segment 4, addresses the reader directly: …create your own "Victorian" 
bouquet of flowers.  Without some indication of why and how the reader should carry out such an 

activity, the BACKGROUND relation provides no support other than comprehension.  Moreover, the 

nucleus of the BACKGROUND is not 4, but the PURPOSE relation itself, containing the structure of 

nested ELABORATION relations (5-8) for which 4 is satellite.  The PURPOSE of 5-8, to create your 

own "Victorian" bouquet of flowers, must stand on its own.  This places 4 at risk of negation.  If 

the reader is disinclined to create a bouquet and denies 4, the PURPOSE relation fails.  Negation of 

the nucleus implies negation of the satellite, and this results in negation of the BACKGROUND 

relation. 

Concerns such as these are opportunities for assessing the structure of the text.  An alternative 

analysis, one that would align the logic of the text and its RST structure, would be to code segment 

4 as the nucleus of a MOTIVATION/ENABLEMENT schema, as shown in Figure 13.  Segments 1-3 

are intended not merely to impart comprehension of what follows, but to increase the reader’s desire 

to perform the action, i.e., to create a bouquet.  By offering the reader the prospect of a full summer 
of floral beauty, segments 1-3 are intended to increase the reader’s desire to create a basket, and 

segments 5-8 increase the reader’s ability to do so, by providing a method involving choices of 

shapes, sizes, forms, and colors of the flowers, height of the plants, leaf textures, and leaf colors.  

With this revised analysis, segment 4 emerges as the locus of effect for the text, and as the logical 

consequent as well.  Without alignment between RST structure with the logic of relational 

propositions, the inferences indicated by the original RST analysis lack plausibility and are subject 

to negation.  By aligning the RST structure with its logical inferences, the organization of the text 
comes into focus: segment 1 follows from text span 2-3, and segment 4 follows from 1-3, 

representing the MOTIVATION part of the text.  Conjunctively, 4 also follows from 5-6, representing 

the ENABLEMENT part of the text.  Here is the MOTIVATION conjunct: 

 

(((((((2  3  (2 → 3)) → 1)  (2  3  (2 → 3))) → 1) → 4)  ((((2  3  (2 → 3)) → 1) 

 (2  3  (2 → 3))) → 1)) → 4) 

 

And this is the ENABLEMENT conjunct: 

 



REASONING BETWEEN THE LINES 

106 

 

(((((5  6  (5 → 6))  (5  (7  8  (7 → 8))  (5 → (7  8  (7 → 8))))) → 4)  ((5  6 

 (5 → 6))  (5  (7  8  (7 → 8))  (5 → (7  8  (7 → 8)))))) → 4) 
 

In their commentary on this analysis, Mann and Thompson seemed to have recognized this, as 

indicated by their observation that 4 presents the possible goal, and that segments 5-8 provide the 

method.  But they did not choose to code the text as such.  An RST analysis that cannot pass the 

test of logic is an analysis in need of attention.  As demonstrated here, the revisions indicated by 

the logic of relational propositions need not be merely a matter of tweaking a few relations.  It may 

provide an alternative perspective on the text.  In this example, the difference is between expository 

and argumentative.  That the logic of relational propositions can be used to assess an existing RST 

analysis shows that it can also be used in the development of new analyses as well.  The decision 

process would be the same: for any structure under consideration, identify the corresponding logical 

expressions and their consequences.  If these lead to inconsistencies, either the text is problematic 

or the structure is in need of review. 

 
Figure 13. Alternative of analysis of Bouquets in a Basket 

6 Conclusion 

The relationship between coherence relations and logic has been an active area of research at least 

since Hobbs (1979, 1985) used logic to define a set of relations for use in an automated inference 

system.  Using this approach, relations could be represented in a manner akin to (but not identical 

with) the predicate calculus (Hobbs, 1979).  There have been numerous other studies in the logic 

of coherence relations (e.g., Danlos, 2008; González & Ribas, 2008; Groenendijk, 2009; Marcu, 

2000; Sanders et al., 1992; Wong, 1986).  Much of the recent research in this area has been 

influenced by the work of Asher and Lascarides (2003), with their development of Segmented 

Discourse Representation Theory (SDRT).  SDRT is based in part on Discourse Representation 

Theory (DRT), a theory of dynamic semantics that attempts to account for the context dependence 

of meaning, based on the observation that how a sentence is interpreted depends on what has 

occurred previously within the discourse (Kamp & Reyle, 1993; Kamp, Van Genabith, & Reyle, 

2011).  That is, the discourse context is dynamically modified with each successive utterance, and 

this updated context informs the interpretation of whatever comes next.  Building on DRT, SDRT 

uses dynamic semantics to generate interpretations of discourse.  Updates to the interpretation are 

made using a glue logic which identifies appropriate rhetorical relations for linking new 

information to old, assigning a compositional and dynamic semantic interpretation for the 

discourse.  The aims of this paper have been more modest than those of SDRT.  In this paper I have 

demonstrated the use of relational propositions as a means for investigating texts as expressions in 

propositional logic.  While SDRT defines a set of rules for deriving rhetorical structures from 



POTTER 

   107 

 

logical forms, the logic of relational propositions provides a set of rules for deriving logical forms 

from rhetorical structures.  While SDRT is a theory of dynamic discourse interpretation, the logic 

of relational propositions is based on a static logical interpretation of intended effect.  Thus, the 

two theories are each motivated by rather different goals, with rather different results. 

This paper has shown that any text analyzable using RST can be restated as an expression in 

propositional logic.  Relational propositions play a key role in this.  By providing a theoretical basis 

for treating RST structures as assertions, relational propositions provide a bridge between RST 

analyses and propositional logic.  In order to apply the logic of relational propositions to complete 

texts, rather than individual relations, it was necessary to extend the theory, so that any multiunit 

RST span can be restated as a nested relational proposition.  For each of the RST relations there is 

a defined logical form reflecting the logical relationship between the satellite and the nucleus.  

Within a nested relational proposition, the forms may be substituted for each of the constituent 

relations.  Mapping the forms into a nested relational proposition yields a logical expression of the 

text.  The reduction of texts to logical forms reveals that logical inference is pervasive within 

coherent text.  The paper has also defined a set of rules for negation of logical forms.  The logical 

forms and rules of negation can be used to examine the logical coherence of texts.  The status of 

the logical coherence of a text can be used to support RST analysis.  Because there is a 

correspondence between logical coherence and the functional relationships of RST,  an RST 

analysis that cannot pass the test of logic is indicative either of a problematic analysis or textual 

incoherence.  

There are several potential applications of this research.  If integrated with computational 

methods for generating RST analyses (e.g., Corston-Oliver, 1998; Hernault, Prendinger, duVerle, 

& Ishizuka, 2010; Pardo, Nunes, & Rino, 2004; Soricut & Marcu, 2003), the method presented 

here could lead to useful tools for scalable analysis of large text collections.  Some other potential 

uses include contributions to knowledge representation, automated reasoning, controlled natural 

languages, cross-document analysis (Cardoso, Jorge, & Pardo, 2015; Radev, 2000), and integration 

with research in semantic equivalence, entailment, and knowledge extraction (Androutsopoulos & 

Malakasiotis, 2010; Gangemi, 2013; Zhang & Patrick, 2005).  Some areas for future study include 

development of a computational framework for the study of the logic of RST analyses, studies in 

the interrelationship between logic and rhetoric, a more in-depth look at multinuclear relations, and 

investigation of various genres of discourse using the methods described in this paper. 

References 

Abelen, E., Redeker, G., & Thompson, S. (1993). The rhetorical structure of US-American and 

Dutch fund-raising letters. Text, 3, 323-350.  

Androutsopoulos, I., & Malakasiotis, P. (2010). A survey of paraphrasing and textual entailment 

methods. Journal of Artificial Intelligence Research, 38, 135-187.  

Asher, N., & Lascarides, A. (2003). Logics of conversation. Cambridge, UK: Cambridge University 

Press. 

Azar, M. (1999). Argumentative text as rhetorical structure: An application of rhetorical structure 

theory. Argumentation, 13(1), 97-114.  

Boole, G. (1854). An investigation of the laws of thought on which are founded the mathematical 

theories of logic and probabilities (Dover 1958 reprint ed.). New York: Macmillan. 

Brooks, A. W. (2013). Get excited: Reappraising pre-performance anxiety as excitement. Journal 

of Experimental Psychology, 143(3), 1144-1158.  

Cardoso, P. C. F., Jorge, M. L. R. C., & Pardo, T. A. S. (2015). Exploring the Rhetorical Structure 

Theory for multi-document summarization Congreso de la Sociedad Española para el 

Procesamiento del Lenguaje Natural (Vol. XXXI). Alicante, Spain: Sociedad Española 

para el Procesamiento del Lenguaje Natural. 



REASONING BETWEEN THE LINES 

108 

 

Carlson, L., & Marcu, D. (2001, September). Discourse tagging reference manual.   Retrieved from 

ftp://ftp.isi.edu/isi-pubs/tr-545.pdf 

Copi, I. M. (1967). Symbolic logic. New York: Macmillan. 

Corston-Oliver, S. H. (1998). Computing representations of the structure of written discourse. 

(dissertation), University of California, Santa Barbara, CA.    

Danlos, L. (2008). Strong generative capacity of RST, SDRT and discourse dependency DAGSs. 

In A. Benz & P. Kühnlein (Eds.), Constraints in discourse (pp. 69–95). Amsterdam: 

Benjamins. 

Forbus, K. D. (2016). Software social organisms: Implications for measuring AI progress. AI 

Magazine, 37(1), 85-90.  

Gangemi, A. (2013). A Comparison of knowledge extraction tools for the Semantic Web. In P. 

Cimiano, O. Corcho, V. Presutti, L. Hollink, & S. Rudolph (Eds.), The Semantic Web: 

Semantics and Big Data (pp. 351-366). Berlin, Heidelberg: Springer. 

González, M., & Ribas, M. (2008). The construction of epistemic space via causal connectives. In 

I. Kecskes & J. Mey (Eds.), Intention, common ground and the egocentric speaker-hearer 

(pp. 127-149). Berlin: de Gruyter. 

Groenendijk, J. (2009). Inquisitive semantics: Two possibilities for disjunction. In P. Bosch, D. 

Gabelaia, & J. Lang (Eds.), Logic, language, and computation (pp. 80-94). Berlin, 

Heidelberg: Springer Berlin Heidelberg. 

Grosz, B., & Sidner, C. (1986). Attention, intentions, and the structure of discourse. Computational 

Linguistics, 12(3), 175-204.  

Halliday, M. A. K., & Hasan, R. (1976). Cohesion in English. London: Longman. 

Hernault, H., Prendinger, H., duVerle, D. A., & Ishizuka, M. (2010). HILDA: A Discourse parser 

using Support Vector Machine classification. Dialogue and Discourse, 1(3), 1-33.  

Heuboeck, A. (2009). Some aspects of coherence, genre and rhetorical structure – and their 

integration in a generic model of text. Language Studies Working Papers, 1, 35-45.  

Hintikka, J. (2007). Socratic epistemology: Explorations of knowledge-seeking by questioning. 

New York: Cambridge University Press. 

Hobbs, J. R. (1979). Coherence and coreference. Cognitive Science, 3, 67-90.  

Hobbs, J. R. (1985). On the coherence and structure of discourse (CSLI-85-37).  Stanford, CA: 

Center for the Study of Language and Information, Stanford University.   Retrieved from 

http://www.isi.edu/~hobbs/ocsd.pdf 

Hovy, E. H. (1990). Parsimonious and profligate approaches to the question of discourse structure 

relations. Proceedings of the Fifth International Workshop on Natural Language 

Generation. Pittsburgh, PA: Association for Computational Linguistics. 

Hovy, E. H., & Maier, E. (1993). Parsimonious or profligate: How many and which discourse 

relations?  Marina del Rey, CA: Information Sciences Institute, University of Southern 

California.   

Iruskieta, M. (2014). A description of pragmatics rhetorical structure and its evaluation in 

computational linguistics. Paper presented at the Programa de Pós-Graduação em Letras, 

Maringa, Brazil.  

Kamp, H., & Reyle, U. (1993). From discourse to logic: Introduction to model-theoretic semantics 

of natural language, formal logic and Discourse Representation Theory. Dordrecht: 

Kluwer. 

Kamp, H., Van Genabith, J., & Reyle, U. (2011). Discourse representation theory. In D. Gabbay & 

F. Guenthner (Eds.), Handbook of philosophical logic, Volume 15 (2nd ed., pp. 125-294). 

Dordrecht: Springer. 

Knott, A., Oberlander, J., O'Donnell, M., & Mellish, C. (2001). Beyond elaboration: The interaction 

of relations and focus in coherent text. In T. Sanders, J. Schilperoord, & W. Spooren (Eds.), 

ftp://ftp.isi.edu/isi-pubs/tr-545.pdf
http://www.isi.edu/~hobbs/ocsd.pdf


POTTER 

   109 

 

Text Representation: Linguistic and Psycholinguistic Aspects (pp. 181-196). Amsterdam: 

John Benjamins. 

Kolodny, N. (2007). How does coherence matter? Proceedings of the Aristotelian Society (Vol. 

107, pp. 229-263). Oxford: Oxford University Press. 

Mann, W. C., & Matthiessen, C. M. I. M. (1990). Functions of language in two frameworks. Marina 

del Rey, CA: Information Sciences Institute. 

Mann, W. C., & Thompson, S. A. (1985). Assertions from discourse structure (Technical Report 

No. ISI/RS-85-155).  Marina del Rey, California: Information Sciences Institute.   

Mann, W. C., & Thompson, S. A. (1986a). Assertions from discourse structure. HLT '86: 

Proceedings of the workshop on strategic computing natural language (pp. 257-270). 

Morristown, NJ: Association for Computational Linguistics. 

Mann, W. C., & Thompson, S. A. (1986b). Relational propositions in discourse. Discourse 

Processes, 9(1), 57-90.  

Mann, W. C., & Thompson, S. A. (1987a). Rhetorical Structure Theory: A framework for the 

analysis of texts. IPRA Papers in Pragmatics, 1, 1-21.  

Mann, W. C., & Thompson, S. A. (1987b). Rhetorical structure theory: A theory of text 

organization (ISI/RS-87-190).  Marina del Rey, CA: University of Southern California, 

Information Sciences Institute (ISI).   

Mann, W. C., & Thompson, S. A. (1988). Rhetorical structure theory: Towards a functional theory 

of text organization. Text, 8(3), 243-281.  

Mann, W. C., & Thompson, S. A. (2000a). Toward a theory of reading between the lines: An 
exploration in discourse structure and implicit communication. Paper presented at the 

Seventh International Pragmatics Conference, Budapest, Hungary. 

Mann, W. C., & Thompson, S. A. (2000b). Two views of rhetorical structure theory. Proceedings 

of the 10th Annual Meeting of the Society for Text and Discourse. Lyon, France. 

Marcu, D. (1996). Distinguishing between coherent and incoherent texts. The Proceedings of the 

Student Conference on Computational Linguistics in Montreal (pp. 136–143). 

Marcu, D. (2000). The theory and practice of discourse parsing and summarization. Cambridge, 

MA: MIT Press. 

Nicholas, N. (1994). Problems in the application of rhetorical structure theory to text generation. 
(masters thesis), University of Melbourne, Melbourne, Australia.    

Pardo, T. A. S., Nunes, M. d. G. V., & Rino, L. H. M. (2004). DiZer: An automatic discourse 

analyzer for Brazilian Portuguese. Advances in Artificial Intelligence – SBIA 2004 17th 
Brazilian Symposium on Artificial Intelligence, Sao Luis, Maranhao, Brazil, September 29-

Ocotber 1, 2004. Proceedings. Berlin: Springer. 

Radev, D. R. (2000). A common theory of information fusion from multiple text sources step one: 

cross-document structure Proceedings of the 1st SIGdial workshop on Discourse and 

dialogue - Volume 10 (pp. 74-83). Hong Kong: Association for Computational Linguistics. 

Rubin, V. L., & Vashchilko, T. (2012). Identification of truth and deception in text: Application of 

vector space model to rhetorical structure theory Proceedings of the Workshop on 
Computational Approaches to Deception Detection (pp. 97-106). Avignon, France: 

Association for Computational Linguistics. 

Russell, B. (1906). The theory of implication. American Journal of Mathematics, 28(2), 159-202  

Sanders, T. J. M., Spooren, W. P. M., & Noordman, L. G. M. (1992). Toward a taxonomy of 

coherence relations. Discourse Processes, 15, 1-35.  

Soricut, R., & Marcu, D. (2003). Sentence level discourse parsing using syntactic and lexical 

information. Proceedings of the 2003 Conference of the North American Chapter of the 

Association for Computational Linguistics on Human Language Technology - Volume 1 

(pp. 149-156). Edmonton, Canada: Association for Computational Linguistics. 



REASONING BETWEEN THE LINES 

110 

 

Sowa, J. F. (2000). Knowledge representation: Logical, philosophical, and computational 
foundations. Pacific Grove, CA: Brooks/Cole. 

Stede, M. (2008). Disambiguating rhetorical structure. Research on Language and Computation, 

6(3), 311-332.  

Stede, M., Taboada, M., & Das, D. (2017). Annotation guidelines for rhetorical structure. 

University of Potsdam and Simon Fraser University.   Retrieved from 

http://www.sfu.ca/~mtaboada/docs/research/RST_Annotation_Guidelines.pdf 

Stent, A. J. (2000). Rhetorical structure in dialog. Paper presented at the 2nd International Natural 

Language Generation Conference (INLG'2000), Patras, Greece. 

Taboada, M. (2006). Discourse markers as signals (or not) of rhetorical relations. Journal of 

Pragmatics, 38(4), 567-592.  

Taboada, M., & Mann, W. C. (2006a). Applications of rhetorical structure theory. Discourse 

Studies, 8(4), 567-588.  

Taboada, M., & Mann, W. C. (2006b). Rhetorical structure theory: Looking back and moving 

ahead. Discourse Studies, 8(3), 423-459.  

Taboada, M., & Renkema, J. (2008). Discourse relations reference corpus: RST analyses of 65 

texts.  Burnaby, Tilburg: Simon Fraser University and Tilburg University.   Retrieved from 

http://www.sfu.ca/rst/06tools/discourse_relations_corpus.html 

Thagard, P. (1989). Extending explanatory coherence. Behavioral and Brain Sciences, 12, 435-

502.  

Thompson, S. A., & Mann, W. C. (1986). A discourse view of concession in written English. 

Proceedings of the Second Annual Meeting of the Pacific Linguistics Conference (pp. 435-

447). Eugene, Oregon: University of Oregon, Eugene. 

Thompson, S. A., & Mann, W. C. (1987). Antithesis: A study in clause combining and discourse 

structure. In R. Steele & T. Threadgold (Eds.), Language Topics: Essays in Honour of 
Michael Halliday, Volume II (pp. 359-381). Amsterdam: John Benjamins. 

Whitehead, A. N., & Russell, B. (1910). Principia mathematica (Vol. 1). Cambridge, England: 

Cambridge University Press. 

Wong, W.-K. C. (1986). A theory of argument coherence (TR86-29).  Austin, Texas: Artificial 

Intelligence Laboratory.   

Zhang, Y., & Patrick, J. (2005). Paraphrase identification by text canonicalization. Proceedings of 

the Australasian Language Technology Workshop (pp. 160-166). Sydney, Australia. 

 

http://www.sfu.ca/~mtaboada/docs/research/RST_Annotation_Guidelines.pdf
http://www.sfu.ca/rst/06tools/discourse_relations_corpus.html

	Reasoning Between the Lines: A Logic of Relational Propositions
	1 Introduction
	2 Theoretical Background
	2.1 Rhetorical Structure Theory
	2.2 Relational Propositions
	2.3 Propositional logic

	3 A Logical Interpretation of Rhetorical Relations
	3.1 Presentational Relations
	3.1.1 Antithesis
	3.1.2 Concession
	3.1.3 Evidence
	3.1.4 Justify
	3.1.5 Motivation and Enablement
	3.1.6 Background
	3.1.7 Preparation
	3.1.8 Restatement and Summary

	3.2 Subject Matter Relations
	3.2.1 Causal Relations
	3.2.1.1 Cause and Result
	3.2.1.2 Means
	3.2.1.3 Condition
	3.2.1.4 Purpose
	3.2.1.5 Unconditional

	3.2.2 Otherwise and Unless
	3.2.3 Solutionhood
	3.2.4 Elaboration
	3.2.5 Evaluation and Interpretation
	3.2.6 Circumstance

	3.3 Multinuclear Relations

	4 Negating Relational Propositions
	5 The Logic of Relational Propositions
	6 Conclusion
	References