Dialnet TheRoleOfCohesiveDevicesAsTextualConstraintsOnRele 713368 PDF

International Journal
?f
Lnglish Siudies
The Role of Cohesive Devices as Textual Constraints

on Relevance: A Discourse-as-process View
ANA 1. MORENO'
University of León
ABSTRACT
Drawing on Relevance Theory (Sperber and Wilson, 1986; Blakemore, 1987, 1992) and taking
insights from Sinclair's (1993) model of written text structure, the purpose of this paper is to
show how and which cohesive features play an important role in helping the reader perceive
relevance and coherence when a text is approached in the process of reading. With this aim, a
comment article from Guurdiun UnIinzited consisting of 60 coherence units is analyzed by a
group of 25 subjects. The study seeks to capture the coherence pattem perceived by a discourse
community rather than by an individual researcher. The results show that in most cases the
cohesive resources that contribute to the perception of the discourse relevance and coherence of
this text at each juncture deal only with discourse meaning derived from whole sentences, larger
fragments of text, or occasionally, certain simple clauses linked paratactically, and they do much
more than effect a tenuous connection between isolated constituents of sentences.
KEYWORDS: discourse, text, reading process, coherence, relevance, cohesive devices,

metadiscourse.
'Addressfor correspondence: Ana l. Moreno, Dpto. Filología Moderna, Facultad de Filosofia y Letras, Universidad
de León, Campus de Vegazana, s/n, 24071 León-Spain, Phone: 987 291095, Fax: 987 291099, E-mail address:
[email protected].
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES, vol. 3 (l), 2003, pp. 11 1-165
1. INTRODUCTION
The present paper focuses on a written act of communication considered from the point of view
of interpretation. Before presenting the theoretical model within which it will be framed, a
convenient distinction will be made and justified between the concepts ofwritten text and written
discourse. Written text will be conceived of as the written record of a potential communicative
event, or one meaningful part of it, where the intended mode of communication between writer
and reader is the written word. Thus, a written text can be anything from a single word written
on a scrap of paper to an entire novel or an academic textbook through a comment article or an
e-mail exchange. In this respect, a written text is seen as the tangible written record of a potential
communicative event.
This definition adds three defining elements as compared to, for instance, Cook's (1989)
general definition of text. On the one hand, the feature of potential to characterize a
communicative evenl is used because a written text is not always eventually read by the intended
reader(s). In the cases where the text is not read at all, it rnay be said that the communicative
event is not fulfilled. In the cases where a non-intended reader reads the text, the communicative
event is not fulfilled exactly as envisaged by the writer. Take, for example, a research article. It
rnay suffer various kinds of fate. Though written with the intention that it should be read by a
wide audience, the text rnay be published or not. If published, it rnay be noticed or not. If
noticed, it inay be found interesting to read. If so, it rnay be read either partially or in its entirety.
In either case, it inay be read by intended andlor by non-intended readers and the number of
readers rnay be very variable. It should be noted that the potential remains with the written word
throughout its legible life, no matter how many times it is read. The point is that until the
potential act of communication made possible through the text is fulfilled, the text-as-product
can only be considered as the written record of an act ofverbal expression rather than the written
record of a communicative event.
This definition also considers as written texts the written records of meaningful parts of
potential communicative events. This is done to include, for instance, written exchanges of non
face-to-face interaction where one or severa1 written replies are expected. Take the case of
certain types of letters or e-mail exchanges. In these cases, it rnay be possible to consider the first
part of the exchange, that is, the initiating letter or message, as one meaningful part of the total
potential communicative event. Finally. the definition excludes from the same set any written
record of a potential coinmunicative event conveyed through a mode other than the written word.
An example of what would be excluded from this definition of written text is the notes for a
paper to be presented orally at a conference, or the lyrics to a song. Although these texts rnay
have been recorded in the written mode, the writer rnay not have conceived them to be decoded
in this mode.
The notion of written discourse, on the other hand, is defined as the meaning perceived
by a reader in the act of interpretation of a given written text at any moment of the interpretation
process. For the last part of this definition at any moment of the interpretationprocess, 1 draw
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES, vol. 3 ( l ) , 2003, pp. 11 1-165
The Role o f lohesive Devices as Textual Constraints on Relevance 113
on Sinclair's (1993: 6) suggestion that "the text at any moment is ... the sentence currently being
interpreted". Just in the same way as structure is necessary in communicating meaning because
we cannot say everything at once (cf. Winter, 1986: 88), when we interpret written discourse we
cannot attend to the whole text at a time. We can only attend to one short stretch of the text at
any time. If a text is seen as a sequence of sentences, the sentence being interpreted is 'the
likeliest unit to cany the status of text oJ'the moment' (Sinclair, 1993: 6). From this perspective,
written discourse can be viewed as a complex unit ofmeaning constantly evolving in the reading
process.
Now, as Sinclair (1993) goes on to mention, in communicating meaning there are two
basic components: that involved in creating meaning and that involved in sharing meaning.
Therefore, language in use "consists in part of features which organise the sharing of meaning,
as well as features which create the meaning" (Sinclair, 1993: 7). This distinction has been
acknowledged by authors such as Vande Kopple (1 985), who, following Williarns' (1 981: 47)
suggestion that in the process of writing "we usually have to write on two levels", developed a
very influential description of metadiscourse. According to Vande Kopple (1985: 83), on one
level we expand propositional content. On the other level, the level of metadiscourse, we do not
add propositional material but help our readers organize, classifi, interpret, evaluate and react
to such material.
The distinction also reminds us of what M.A.K. Halliday has shown repeatedly in his
work, that when we use language, we nearly always work toward fulfilling the three macro-
functions of language. That is, we try to give expression to our experience, to interact with our
audience, and to organize our expression into a cohesive and coherent text. In other words,
Halliday (1973) asserts that we try to convey what are essentially three different kinds of
meaning, which he calls ideational, interpersonal, and textual. The ideational elements could also
be called representational or informational. Interpersonal meanings would allow us to reveal our
personalities, to evaluate and react to the ideational material, and would also include forms of
interaction and social interplay with other participants in the communicative event. Textual
elements would enable the speaker or writer to organize what he is saying in such a way that it
makes sense in the context and fulfils its function as a message (cf. Vande Kopple, 1985: 86).
Each text is an integrated expression of these three kinds of meanings.
Drawing on Sinclair (1993: 7), 1would like to show how, both in their interpersonal and
textual function, metadiscourse elements are part of the interactive apparatus of the language,
that progressively determine the status of previous text in relation to the current sentence. They
both serve to give independence to the sentence and help to perceive it as relevant. However, for
obvious reasons of space limitation, the focus will be narrowed down to one set of elements
within the textual component of metadiscourse: those approximately corresponding in function
to the cohesive devices acknowledged by Halliday and Hasan (1976): reference, substitution,
ellipsis, conjunction and lexical cohesion. This will leave aside metadiscourse elements of the
interpersonal type.
Q Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 (l), 2003, pp. 1 1 1-165
114 Ano l. Moreno
A lot of attention has been paid to cohesive devices, especially since Halliday and Hasan
(1976) described them in the most comprehensive and widely available account. Since then, most
models of cohesion in English have attempted to account for the explicit linguistic devices used
in texts to signal relations between sentences. Their main objective has been to explain the
principles that govem the well formedness, the unity and connectivity of texts by looking at the
different kinds of ties established within texts and the relations they express. However, stating
that the mere repetition of items in texts, or the use of synonyms or superordinates, and so on,
contribute to our perception of the text as coherent does not seem very convincing. In addition,
in most of these models texts seem to have been approached as products rather than processes.
In other words, analysts seem to have treated texts as objects of research, which could be read
and reread as many times as necessary in order to identi@ the different types of cohesive tie.
The position that this paper takes, following Sinclair (1992, 1993), differs from the
previously mentioned views of cohesive items in that, instead of emphasizing the role of these
elements in a text analyzed by the researcher as a finished product, the focus will be on their role
as elements of the interactive apparatus of the language in the process of interpretation of the
text. The major reason for changing the perspective is that ordinary users of the language are
more likely to approach texts as processes rather than as products. Although ordinary users of
the language are likely to be unaware of the kind of analysis reported here, researching this other
side of the coin could be more relevant for communicative participants other than linguists or
would-be linguists.
The model proposed here advocates a study of the role of these elements in the perception
of relevance and, therefore, coherence in the process of interpretation of discourse. Its ultimate
purpose is to determine which textual features of a given text are more likely to help potential
readers to make sense of a discourse-as-process. That is, the present study will try to identi@
those textual elements that help readers to achieve optimal relevance (cf. Sperber & Wilson,
1986) at each successive text of the moment in relation to the growing meaning derived from
processing previous text. They will then be accounted for as textual constraints on relevance, that
is, as test pointers that help readers to select relevant contextual assumptions brought to bear on
the interpretation of current discourse. A sentence will be said to be relevant if it conveys
relevant information and relevance will be defined, following Blakemore (1987: 1 1 l), in terms
of a relationship between propositions.
This study will then attempt to contribute to answering the intriguing question of what
helps readers to perceive a piece of text as coherent. If coherence at a given point in a text is
understood as a relation between linguistic units (Blakemore, 1987: 11 l), then being able to
perceive the relevance of a text segment at that point in the process of reading may contribute
to perceiving the text as coherent at that point. In connection with this aim, 1 would like to
confirm Sinclair's (1993) suggestion that only a relatively small number of the textual features
normally accounted for in cohesion analyses play a crucial role in helping the reader both
perceive the relevance of each successive sentence, or text of the moment, and eventually, the
Q Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 (l), 2003, pp. 1 1 1-165
The Role of Cohesive Devices as Textual Constraints on Relevante 115
coherence of the whole text. 1 believe that this may have important implications, for instance,
for the design of reading and testing tasks that focus on this aspect of text comprehension (cf.
Moreno 1998a). However, the type of research reported in the present paper will in principle
differ from Sinclair's (1993) and Moreno's (1 998a) in that the study will attempt to capture the
coherence pattem perceived from a text by a group of individuals rather than by an individual
readerlresearcher.
11. THEORETICAL FRAMEWORK

The model of text structure proposed by Sinclair (1 993) is based on the assumption that when
the reader moves on to processing the following sentence and focuses on its interpretation, the
linguistic properties ofthe previous sentence are discarded and only what it expreses is retained.
"It is no longer a linguistic entity, but a part of shared knowledge" (Sinclair, 1993: 9). Hence the
present distinction between the text, the tangible linguistic object, and the discourse, the meaning
generated from interpreting the text, seems to be consistent with Sinclair's view of the reading
process.
If we agree that a text can be qualified as coherent when it is perceived as unified and
meaningful to a particular reader, then it can be said that a written text achieves ail its potentiai
of unity and meaningfulness, that is, ail its potential of being perceived as coherent, when the
process of interpretation reaches its end, as envisaged by the writer. However, readers do not
need to wait until they have finished reading the whole written product to try and make sense of
the text. Competent readers will attempt to make sense of the discourse from the very moment
the reading process begins, and -if motivation and interest endures- may continue doing so
at every stage in the reading process. In other words, accomplished readers will attempt to
retrieve discourse meaning as they come across subsequent textual units in their search for
relevance.
The idea of relevance was first proposed by Grice (1975: 46) as one of the four maxims of the
cooperative principle, a general principle of communication known and applied by al1 human
beings. Although Grice used his theory mainly to account for the language of conversation, the
basic tenet of his theory may be adapted to account for what happens in the process of
interpretation of written language in such a way that we can assert that readers expect texts to
be a co-operative effort and to progress in a rational manner. And if texts are to be rationai, they
must consist of sentences that are in some way connected to each other. What guarantees this
connection is what he calls the co-operariveprinciple, and one of the four maxims that comprise
this principle is the maxim of relation: be relevant.
One problem with Grice's theory is that it gives no indication of how speakers may be
relevant, that is, of how a given utterance is perceived as relevant by the hearer. Sperber and
Wilson (1986) attempt to answer this question in a more detailed way. They argue that an
explanation of utterance interpretation must be based on a general cognitive theory of
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 ( l ) , 2003, pp. 1 1 1-165
-
116 Ana l. Moreno
information processing. For the same reason, it should be posible to consider an explanation of
sentence interpretation (the equivalent minimal unit of coherence in written language) from the
same general theory. Thus 1will be using the terms reader, writer and sentence in places where
authors such as Sperber and Wilson (1986) and Blakemore (1 992) would use hearer, speaker,
and utterance respectively. The basic idea underlying their principle of relevance is that in
processing information, people generally aim to bring about the greatest improvement to their
overall representation of the world for the least cost in processing. Following this line of
reasoning, we could say that readers go ahead and interpret every sentence in the expectation that
it will interact with their existing assumptions to yield a contextual effect (i.e. a new
assumption).
A contextual effect is defined as the impact of a new item of information on an existing
representation of the world. Sperber and Wilson (1 986) identifj three ways in which a new item
of information rnay have a contextual effect: 1) It rnay allow the derivation of a contextual
implication, thereby allowing readers to add assumptions to their existing representation of the
world; 2) It rnay provide further evidence for, and hence strengthen, an existing assumption; and
3) It rnay contradict an existing assumption. In this latter case, the reader rnay decide to abandon
the existing assumption in favour of the information that has been presented to him. When an
item of information has a contextual effect in a given context, Sperber and Wilson say it is
relevant in that context.
For the notion of context, 1 draw on Blakemore's (1992) view of context as a
psychological construct. According to her definition, derived from Sperber and Wilson's (1 986)
theory of relevance, the context is the set of assumptions that are brought to bear in interpreting
a given sentence. Contextual assurnptions rnay be derived from different sources: from direct
observation (i.e. from the situational context), through the interpretation of the preceding text
(i.e. from the co-text), or from information stored in memory (i.e. fiom background knowledge,
be it cultural or interpersonal). On processing a new item of information the role of contextual
assumptions would be to combine with the assurnption derived from interpreting the sentence
in question (or text ofthe moment) as premises in an argument. Thus establishing the relevance
of the new assumption would involve inference. It would involve the interaction of existing
assumptions with the new assumption. Therefore, the relevance of a new assumption would
depend on the context in which it is processed.
One important question is which principle readers follow to choose the particular
contextual assumptions they bring to bear on the interpretation of a new sentence. For, logically
speaking, any of their beliefs and assumptions rnay be brought to bear. In answer to this
question, Sperber & Wilson (1986) argue that intuitively, it is clear that the greater the impact
a proposition has on the readers' representation of the world the greater its relevance. Also,
accessing contextual assumptions and using them to derive contextual effects involves a cost,
and the cost of deriving contextual effects in a small, easily accessible context will be less than
the cost of obtaining them from a larger, less accessible context. Of course, there are other
O Servicio de Publicaciones.Universidad de Murcia. All rights reserved. IJES, vol. 3 ( l ) , 2003, pp. 1 1 1-165
Thr Rolr ofCohesivr Drvicrs os Textual Constroints un Relrvonce 117
factors that may also affect the effort made in achieving relevance, such as the readers'
interpreting ability, their leve1 of concentration, their interest, their emotional states, and so on.
Let us assume for a moment that we are dealing with very motivated competent readers. Thus,
readers who are searching for relevance at every text of the moment will process each new item
of information in the context that yields a maximum contextual effect for the minimum cost in
processing. According to Sperber and Wilson (1986), what the writer does is manipulate the
readers' search for relevance. Obviously, it is in the interest of these readers that the writer
should produce a sentence whose interpretation calls for less processing effort than any other
sentence that s/he could have made. But equally, given that writers wish to communicate with
potential readers, it is in their interest to make their sentences as easily understood as possible.
This means that readers are entitled to interpret every sentence on the assumption that writers
have tried to give them adequate contextual effects for the minirnum necessary processing, or,
in other words, that writers have aimed at optimal relevance.
In written text, the most easily accessible contextual assumptions are those derived from
the preceding text, and in particular those still stored in the readers' short-term memory. As
Blakemore (1987: 112) suggests, the most easily accessible assumption at any stage of the
reading process is likely to be the one derived from the immediately preceding sentence. This
assumption made by readers and writers in their cooperative effort to communicate facilitates
the interpretation process in such a way that readers can assume that the new sentence should
be interpreted at least in the context provided by the assumption derived from the preceding
sentence, which in its turn has been interpreted in the context generated from interpreting its
preceding sentence, and so on. Orto put it in a way parallel to Sinclair's (1 993) terms, we might
say that readers can assume that the text of the moment can be interpreted at least in relation to
the meaning derived from the text of thepast. In this sense we may say that readers can trust the
text of the past to a great extent in order to achieve relevance at any text of the moment (cf.
Sinclair, 1992). 1say to a great extent because readers, of course, may and actually do use other
sources of meaning as well as that derived from previous text (see above).
The picture of discourse which. according to Blakemore (1 987: 122), emerges from this
relevance-based frarnework is one in which the interpretation of a sentence (that is, its
propositional content and its contextual effects) contributes toward the context for interpreting
subsequent sentences. That is, as discourse proceeds, readers are provided with a gradually
changing background against which new information is processed. As Blakemore explains,
interpreting a sentence involves more than identiQing the proposition it expreses. It also
involves working out the consequences of adding it to the readers' existing assumptions, or, in
other words, working out its relevance (cf. increment, Brazil, 1995).
In Blakemore's view, two utterances may be connected in coherent discourse in either
of two ways:
Either in virtue of the fact that the interpretation of the first may include propositions used in
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES, vol. 3 ( l ) , 2003, pp. 1 1 1-165
establishing the relevance ofthe second, or in virtue of the fact that a proposition conveyed by one
is affected by the interpretation of the other. In either case we might say that the relevance ofone
is somehow dependent on the interpretation of the other.
Blakrmorr (1987: 122)
Given the role of inference in establishing the contextual effects of a proposition,

Blakemore (1987) suggests that it should not be surprising that expressions that instruct the
reader to establish an inferential connection between two segments of discourse rnay be used to
indicate how the proposition they introduce is to be interpreted as relevant. Thus expressions like
so, ajter al1 and moreover can be used to express relationships of dependent relevance. In this
sense, expressions such as these can be considered as semantic constraints on relevance. But not
only do these expressions have a role in pragmatic interpretation (in helping to perceive
dependent relevance), their role is also important in helping the writer optimize relevance in
accordance with the Principle of Relevance. In other words, they allow writers to make sure that
their readers select the most effective contextual assumptions in their search for relevance at
minimal processing cost.
In this sense, cohesive mechanisms, the object of the present research, can be considered
as constraints on relevance for both types of coherence. This paper will look at the role of
cohesive devices both in establishing the relevance of a new sentence and in helping readers to
retrieve the most cost-effective contextual assumptions in the actual process of reading. Given
this framework, the present study will attempt to determine which cohesive mechanisms of a
given text are more likely to help potential readers to make sense of a discourse-as-process, i.e.
to perceive the relevance of each new text of the moment and ultimately to perceive the
coherence of the entire text.
111. DESIGN OF THE EMPIRICAL RESEARCH

Just as a text-as-product can experience various fates, there rnay be different reasons for wanting
to process a text, i.e. for retrieving discourse meaning, and, therefore, for improving one's
representation of the world. For example, a text can be processed for pleasure, to kill the time:
or for the sake of leaming new things. On the other hand, a text rnay be read for practica1
reasons, whether for personal or for work purposes (e.g. to be able to plan an evening out, to
know how to cook a dish, to be able to repair some technical apparatus). A text rnay also be read
for social reasons (e.g. to have something to talk about). Or it rnay be processed for study
purposes (e.g. to learn about a given field and be able to show one's improved knowledge about
it). Another possible reason for approaching a text is for research purposes (e.g. when the text
is processed by a linguist as an object of analysis). In addition, a text can be read for its general
ideas or to find specific information (cf. Nuttall, 1982: 2).
The important implication from there being such a variety of reasons for reading a text
is that there rnay be multiple resulting discourses even from reading the same text, depending
on the reader's motivation to read it. Furthermore, even for readers with similar purposes,
interpretation of the same text rnay vary depending on the context in which the text is processed
-not only on the assumptions derived from interpreting the previous discourse but also on other
assumptions derived from memory (e.g. previous knowledge of the subject matter) and the other
factors mentioned above.
Thus, if so many different discourses can be derived from a single text, studying the
textual features more likely to contribute to perceiving a text unit as relevant seems an almost
impossible task for at least two main reasons: 1) Because the perception of this property is
ultimately subjective. That is, a text unit rnay be meaningful -and thus rnay contribute to
improving one person's representation of the world-in a way that another person does not have
the necessary knowledge or leve1 of concentration to make sense of. 2) Because the perception
of this property depends on the reader's ultimate aim in reading the text. In other words, a reader
rnay not have perceived the relevance of a coherence unit or the coherence of the whole text
because this rnay not have been the reason for approaching the text.
In addition, it is well known that there is more to the perception of relevance than the
presence of explicit cohesive devices in a text. That is, there are other factors that contribute to
this perception which have nothing to do with textual features. Take, for instance, a reader's
assumptions about the likely sequences of discourse functions derived from memory and
acquired on the basis of previous discourse processing experience. In a situation in which every
single reader will have had different experiences of reading, it does not seem sensible even to
attempt to research the question. However, being aware of the fact that the role of cohesive
devices in perceiving relevance is partial, it rnay be worthwhile investigating which of these
textual features are more likely to contribute to our perception of relevance, if we take certain
precautions. In order to design the most appropriate research strategy for this study, a number
of considerations about the reading process were taken into account.
111.1. Research strategy

One problem derived from the subjectivity factor is that measuring the relevance and coherence
of a piece of discourse is a very difficult task. The first question that rapidly arose is the
following. When can we say that a person has made sense of a text? In relation to the writer's
intentions. to an intended reader's interpretation, to a non-intended reader's interpretation, to a
researcher's interpretation, to some standard of coherence? For the purposes ofthis research, the
most adequate answer seemed to be the last altemative. As Cook (1989: 7) points out, "in
practice we find that discourse is usually perceived as such by groups, rather than individuals".
However, this solution posed a further problem: How could a standard of coherence be
established? The solution given to this problem in the present study was to assume that the
coherence of a text could be established by abstracting away from the particular appreciation of
any individual subject (including the researcher herself) to capture instead the pattern of
120 Ana I. ,\,fo~.eno
coherence perceived by the majority of a group of individuals. Consequently, the present study
attempted to capture the standard of coherence of a given authentic written text as perceived by
a group of individuals in their communicative role as readers of the same text.
111.2. The subjects

It also seemed sensible for this study to control for certain factors which are known to determine
the discourse meaning actually derived from a text. One such factor was the receiver or the type
of reader, with al1 the factors that this may bring along: previous knowledge of the subject
matter, previous reading experience, language proficiency leve], motivation and so on. Thus the
present study focussed on a well-defined group of readers: the discourse community of Spanish
advanced learners of English at the University of León interested to some extent in the
description of the English Language.
Two groups of learners were used in the study. The first group was made up of seven
doctoral students taking a course in Cohesion in English, which was used as part of the pilot
study on which the final study was designed. Actually, one of these students wrote a paper
discussing his interpretation of the text (Pérez Álvarez, 2001). The final study was carried out
with a second group of 25 undergraduate students taking a course in Contemporary Descriptive
Models ofEnglish (taught in the third year of English Philology but also taken by students in the
fourth year). Both groups were characterized as potentially non-intended readers of the targeted
text. Both groups had been provided with a short introduction to the role and type of cohesive
devices based on both Halliday and Hasan's (1976) account and Sinclair's (1993) view of
cohesive devices, using examples from a variety of sources.
The reason for choosing this type of reader lies in the researcher's interest in making the
results of the present study applicable to English as a foreign language reading situations, both
in English for Specific Purposes and General English. As is frequently the case with many of the
texts now used in most EFL learning and assessment situations, learners are required to approach
authentic texts that were not in principle intended for them. Thus the study results both in
relation to the pattern of coherence perceived and the textual elements that help to perceive it
might be extrapolated to readers of this kind but not to others for the reasons argued above.
111.3. The data

The selected text, drawn from the Internet, is a comment article from Guardian Unlimited
consisting of 60 coherent units (56 sentences) and representing typical argumentative written
text. The appendix offers a segmented version of the text. The original version can be found on
the website provided in the references section (Moriarty, 1999). The reader of this paper is
recommended to read the whole text before starting to read the following section. It should be
noted that the text elements presented in parentheses and preceded by an asterisk were not part
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES,vol.3(1),2003,pp. 111-165

The Role of Cohesivr Drvices as Trrtuul Constraints on Relevante 121
of the original text.

The reasons for choosing this text lay both in its topic and its length. On the one hand,
the topic of the article, "a pilot scheme consisting in returning examination papers to the original
candidates", was relevant in the context of the teaching unit on Education that the subjects were
experiencing in one of their courses, Lengua Inglesa 111, at the time the empirical work was
being carried out. And it was one of the texts that the students in this course would have to
summarize. This way, the study would not interfere unnecessarily with their courses and students
would find the kind of analysis done useful for their academic purposes. On the other hand, since
the study would have to be based on only one complete text, the selected article had to meet at
least two requirements. It had to be long enough to yield sufficient material for analysis. But
more importantly, it had to be short enough so as to be manageable for an analysis as complex
as the one proposed here with student subjects.
111.4. The procedure: a discourse-as-process view of the text

Since another determining factor in perceiving text coherence may be the purpose for reading
the text, the present paper also attempted to control for this factor. This was achieved by setting
up a clear expected outcome for the subjects to produce as a result of the reading task. The
required outcome in this case was a written summary of the text. Having to carry out this task
was considered as an incentive for the reader to understand the text as best as possible at every
point. It was assumed that understanding the text well would involve appreciating the
relationship between each new sentence and previous discourse. This skill was considered
crucial in establishing a hierarchy of text units so necessary for summarizing the text, i.e. for
distinguishing main points from secondary details. It was assumed that in order for the subjects
to perceive these relationships, they would have to read the text in great detail.
Obviously, in order to complete the summary task, the subjects would have to read the
text various times. However, for the purposes of this research, only the interpretation taking
place during the first detailed reading of the text was taken into account. Setting up the task of
summarizing the text was then only used by the research design as a stimulus for a detailed
reading that would make it possible to control for the confounding factor ofpurpose ofreading.
Another confounding factor was the difficulty of the text in terms of vocabulary and structures.
To control for this factor the subjects were provided with a glossary with the most predictably
difficult items.
The procedure for obtaining the data was the following. The subjects were first presented
with the original text so that they could have access to its visual features. The title and subtitle
of the text, considered as important elements of the text in generating expectations about the
content and purpose of the whole article, were discussed with the subjects. However, since these
two units seemed to stand on a different hierarchical leve1 in relation to the core text, they were
excluded from detailed analysis. Thus it was the first sentence of the core text that was taken as
122 Ana l. ~tforeno
the first unit of coherence for focussed attention.

Then the subjects were asked to do a number of interpretation tasks. To make sure al1
subjects focussed on intersentential relations rather than intrasentential relations (cf. Hyde, 1990,
2002; Moreno, 1997, 1998b). the text was split up into its constituent coherence units. In most
cases. the minimal unit of coherence corresponded with the orthographic sentence, or the clause
complex (Downing and Locke, 1992), enclosed by a full stop. However, based on Sinclair's
(1993) conclusions about this issue. a few variations were introduced. Sinclair identified a
number of interna1 acts of reference "which may suggest that we revise the original assumption
that the orthographic sentence is the best minimal unit for text structure. In the text-order
analysis, the sentences we choose to divide are those whose two parts behave as two separate
sentences in terms of this analysis. This is a circular argument, but a satisfying one nevertheless.
That is, we do not make arbitrary or intuitive divisions of sentences" (Sinclair, 1993: 19).
If we look at the clauses where Sinclair identifies interna1 acts of reference, they seem
to be enclosed by a colon, a dash, or a comma followed by a coordinator and a reference item,
o r a quotation mark (Sinclair, 1993: 25-28). And it is true that these clauses seem to behave quite
independently from their neighboring clause from the point of view of coherence. In fact, as
Downing and Locke (1992: 283) recognize, "it may be difficult in the spoken language, ..., to
decide whether such combinations of clauses can be considered to constitute a clause complex,
that is to say a single unit, or whether they are to be interpreted as two separate clauses". And
this may also apply to written language. Hence, the question remains, should we continue
considering the pure orthographic sentence as the minimal unit of coherence or should we
reconsider this concept? Sinclair's results could be taken as suggesting that certain types of
paratactic relations between clauses are used by writers to present chunks of information as quite
independent from the point of view of coherence. However, this still needs to be elucidated.
For the purposes of the present study, in order not to make arbitrary or intuitive divisions
of clauses within the different clause complexes in the text, it was decided to divide sentences
at points where there was a colon (1, 17, 28, 41, 47, 58), a dash (20,34), or a comma or dash
followed by some cohesive device (3,29,3 1,56 and 59), provided the following unit could stand
as independent from a coherence point of view. No divisions between clauses in hypotactic or
dependent relationship were made. This is the only place where the researcher had to impose her
own interpretation of what could be considered as an autonomous unit from the point of view
of coherence beforehand. However, this imposition was necessary to guarantee the validity of
the results, i.e. to guarantee that al1 the subjects were observing the phenomena that the study
was focusing on. Arriving at a consensus on this aspect would have constituted another s t ~ d in y
its own right. It should be recalled that the text elements presented in parentheses and preceded
by an asterisk in the segmented text were not part of the original text but are meant to represent
the type of connection inferred by the sub.jects between text fragments (cf. appendix).
As the effort made by each subject in their search for relevance at every text oj the
moment might also be a source of variation, al1 subjects were required to make their
The Role oflohesive Devices as Textual Constraints on Relwance 123
interpretations explicit at every stage in the text in the form of a test as illustrated below. The test
asked them to read the first coherence unit in the text, (l), and then stop reading to try and see
if they perceived a connection between this unit and its co-text. It was assumed that the first
coherent unit would provide the most relevant assumption for interpreting the second coherence
unit. Therefore in test item ( 1 ) students would only perceive a connection between coherence
unit (1) and its co-text if that coherence unit were establishing a prospection. From test item (2)
onwards. al1 types of connections were possible. both backwards and fonvards.
Table 1: Sample Test Items (1 to 2)
Nineteen ninety-nine was the year we dipped a toe in the water:

A) Can you perceive any connection between coherence unit (1) and its co-text? Yes O No
F) Does coherence unit (1) lead you to expect something specific in the following text? Yes O No O
G) lf this connection is explicit. circle and write down (the) prospective signal(s) that make(s) it explicit:
and you know what?

A) Can you perceive any connection between coherence unit (2) and its co-text? Yes O No O
B) lfyou perceive an explicit connection with previous text, circleand write down the retrospective signal(s) that
make(s) it explicit:
C) If you perceive an implicit connection, provide a signalitext fragment to make it explicit:
D) ln relation to which part of previous text can you perceive this connection, whether implicit or explicit?
A O (a word) B O (a phrase) C O (a clause) D1 7 (a sentence) E O (a larger unit)
E) In which sentence(s) is that part of previous text? N o p
F) Does coherence unit (2) lead you lo expect something specific in the following text? Yes O No O
G) lf this connection is explicit, circle and write down (the) prospective signal(s) that make(s) it explicit:
H) Does coherence unit (2) satisfy a prospection created in previous text? Yes O No O
1) If so, in which coherence unit was the prospection created? No
The aim of this test was to guide the subjects' interpretation process at each state of the
text and to help them identify the text features of the current sentence that contributed to their
achieving relevance, i.e. to their perceiving a connection of some kind in relation to previous or
upcoming text at that point in the reading process. Wherever the connection between consecutive
text fragments was not explicit, the subjects were questioned about the type of connection that
helped them to perceive the relevance of the text at that point by making it explicit themselves
by some text item. Subjects were also asked to indicate which part of the text provided them with
the most cost-effective contextual assumptions to establish the relevance of the new proposition.
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 ( 1 ), 2003, pp. 1 1 1 - 165
124 Ana l. ,tforeno
Thus the test was oriented to helping the students capture the following possible types of
phenomena in each current sentence, either retrospective +ncapsulating (anaphorict
elements, (questions B to E), or prospective -predictive ( c a t a p h o r i c t elements (questions F
to G). These two phenomena had been amply acknowledged in the literature (cf. Halliday and
Hasan 1976; Hyde 1990,2002; Sinclair 1993; Tadros 1985,1994). The subjects were also asked
to observe whether the current sentence fulfilled a prospection created in a previous state of the
text (questions H to 1) (cf. Sinclair, 1993). To avoid giving any specific clues. they were asked
to answer the same nine questions (Ato 1) about each new coherence unit in the text, except for
unit (1) at which point only questions A) F) and G), about prospecting mechanisms, were
relevant.
The reason for carrying out the process this way had to do with ensuring that every
individual had made an effort to interpret the current coherence unit before attempting to
interpret the next. Once every individual had specified their interpretations in writing, a round
of discussions was opened to contrast the different interpretations, first in groups of five
individuals made up in such a way that al1 groups were balanced in terms of proficiency level.
Then the discussion was held open-class. The subjects were given the chance to modify their
interpretation after each round of discussions. The aim was to find out whether it was possible
to arrive at a consensus on the most likely interpretation at each stage of the text. It should be
recalled that the main objective was to capture the coherence pattern of the text as perceived by
the group rather than by each individual.
IV. METHOD OF ANALYSIS

Once the data had been gathered following the previously mentioned procedure, it was the
researcher's task to arrange and classify them. The method of analysis consisted first in
distinguishing explicit from implicit relations. Then it analyzed both the text features identified
in the text and those inferred by the subjects. Let us now focus on the major types and subtypes
of coherence mechanisms.
IV.1. Encapsulating, or retrospective, mechanisms (E)

According to Sinclair (1 993), encapsulating mechanisms are those text features identified in the
new sentence that somehow refer back to the meaning created by the whole of the previous
sentence. "By referring to the whole of the previous sentence a new sentence uses it as part of
the subject matter. This removes its discourse function, leaving only the meaning which it has
created" (Sinclair, 1993: 7). It is in relation to this area where clear links could be established
between Blakemore's (1 987) view of relevance and coherence, and Sinclair's (1993) account of
the encapsulating mechanisms used in perceiving text structure, and therefore, coherence. As
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES, vol. 3 (l),2003, pp. 1 1 1-165
The Role of Cuhesive Devices us Textlrul Construints un Relevunce 125
will be recalled, Blakemore (1987) states that two utterances may be connected in coherent
discourse in either of two ways.
Ir 1.I . Relevance of content

One type of coherence arises when infornlation made available by the interpretation of one
segment of discourse is used in establishing the propositional content of the next (Blakemore,
1987: 112). 1propose to call this type relevance of content to distinguish it from what 1will refer
to as relevance of relationa1,function (see below).
IV. l . 1.a. Encapsulating deictic act

Relevance of content closely relates to what Sinclair (1993) terms deictic acts, which certainly
is the area where the mechanism of encapsulation is better appreciated. From the examples he
includes under this category (Sinclair, 1993: 11- 12), it is possible to infer that deictic acts include
phenomena such as reference items and lexical cohesive items, sometimes used in combination.
Let me show this by means of an example taken from the text analyzed, where the numbering
in brackets shows the corresponding coherence units.
(39) In my day, 1 was expected to annotate scripts to explain my marks to the chief
examiner. (40) Remove that requirement, and the examining process will only appear to
be more open, while in fact retaining an almost smug inscrutability.
It is clear that there is one segment in the second sentence, that requirement, whose
interpretation is affected by the interpretation of another segment of previous discourse. In other
words, we can say that in order to establish part of the content of the second proposition we need
to use the propositional meaning created by the interpretation of previous text. That is to say, we
need to bring to bear an assumption derived from having interpreted a previous segment of text.
And, in this case, the segment of previous text that provides the reader with the most relevant
assumption, derived from interpreting semantic content, is the previous sentence: i.e. the
requirement that in her day, the author was expected to annotate scripts to explain her marks to
the chief examiner.
Thus this example shows quite clearly that the encapsulated text is the whole of the
previous sentence. However, not al1 examples of relevance of content seem so clear as this. In
fact, Sinclair (1993) opens an interesting debate that is especially relevant in the two areas of
cohesion included in his framework under the category of deictic acts, namely, reference and
lexical cohesion. The debate refers to a possible distinction between the process of encapsulation
and what Sinclair identifies as point-to-point cohesion. According to Sinclair (1 993), there are
other kinds of cohesion that refer to less than a sentence, and these are not regarded as textual
in nature. To clarify this distinction terminologically speaking, 1 propose to call the process of
true encapsulation textual cohesion, as opposed to point-to-point cohesion.
126 Ana I. 1if01,eno
Point-to-point cohesion may be said to occur in cases where a pronoun can be related
back to a noun phrase earlier in the text, and can be said to refer to it. This kind of pattern is
clearly of frequent occurrence, and is the basis of most accounts of cohesion and the focus of
many tasks in teaching reading discourse skills. It includes the rich field of lexical cohesion,
where the recurrente of a word or phrase, or the occurrence of something reminiscent of a
previous item is noted. Each constituent of these pattems is less than one sentence long;
normally a word or phrase, or at most a clause (Sinclair, 1993: 8).
Since one of the first studies where this type of items was discussed at length is Halliday
and Hasan (1 976), let me show one simple and trivial example that they take from a cookery book
to illustrate what Sinclair seems to callpoint-to-point cohesion:
[ l ] Wash and core six cooking apples. Put them into the fireproof dish.
According to Halliday and Hasan (1976: 3), "it is clear that them in the second sentence
refers back to (is anaphoric to) the six cooking apples in the first sentence. This anaphoric function
of them gives cohesion to the two sentences, so that we interpret them as a whole; the two sentences
together constitute a text. Or rather, they form part of the same text since there may be more of it
to follow". And later they go on to say that "the meaning of the cohesive relation between them and
six cooking apples is that they refer to the same thing. The two items are identical in reference, or
co-referential. So, in this case, what provides the texture is the co-referentiality of them and six
cooking apples". (Halliday & Hasan, 1976: 3). This is the way the authors analyze most cases of
cohesion, other than conjunction:
... where the cohesive item is something like he or une, which coheres by direct reference to, or
substitution for, another item the presupposed element is typically a specific item in the
immediately preceding sentence. [...]Characteristically these instantes also tend to form cohesive
chains, sequences in which it, for example, refers back to the immediately preceding sentence
-but to another it in that sentence, and it is necessary to go back three, four or more sentences,
stepping across a whole sequence of its, before finding the substantial element.
Halliday B Hasan (1 976: 15)
As has been noted, in Sinclair's (1 993) view, these cases of cohesion are not regarded as textual
in nature. According to him, textual cohesion deals "only with sentences or, occasionally, clause
complexes, or even longer stretches of text, and it does much more than effect a tenuous
connection between isolated constituents of sentences. It is the process of encapsulation, and it
reclassifies a previous sentence or text by demoting it into an element of the structure of the new
sentence" (Sinclair, 1993: 9). A clear example of the process of encapsulation has been discussed
at the top of this section (39-40), where it was easy to observe how the meaning of the first
sentence, (39), had been demoted into the direct object of the first clause in sentence (40).
The Rale o/ lohesive Devices as Textual Conshaints un Relevante 127
As Sinclair (1993: 8) claims, "failure to appreciate the distinction between these two
types of cohesion has hampered the development of models of text structure". The model of text
that he puts forward "has no place for retention of the actual words and phrases of text so that
such connections between text items could be established" (Sinclair, 1993: 8). The model 1
advocate also adopts the same perspective. However, it reconsiders some cases that might be
considered aspoint-to-point cohesion by Sinclair.
Let us take fragment [ l ] again:
[ l ] Wash and core sir cooking upples. Put them into the fireproof dish.
In my view, it is possible to appreciate a mechanism of encapsulation as defined by

Sinclair, Le., by perceiving a cohesive relation between them and the meaning created by
interpreting the whole of the previous sentence. The best way to see this is by trying to visualize
the state in which the sir cooking apples that them refers back to are when we put them into the
fireproof dish: are they the same six cooking apples that we originally took to be washed and
cored or are they six cooking apples already washed and cored? When the group of 25 subjects
were asked this question during their short course they unanimously answered that the six
cooking apples that we need to put into the fireproof dish, according to the recipe, should already
be washed and cored.
Thus it is possible to suggest that in this case the relevance is provided in this respect by
the fact that the interpretation of them is affected by the interpretation of the semantic content
of the whole previous sentence, not only by the interpretation of the phrase sir cooking apples,
as Halliday and Hasan suggest. Thus, this case of them would be analyzed in our methodological
framework as a case of textual cohesion because it encapsulates the meaning created by the
whole of the previous sentence.
This departure from previous analyses pertains specifically to cases of cohesion such as
the one illustrated by [l], where reference occurs, and some cases of lexical cohesion. If we
relate this position to the relevance-based framework, it may be said that the personal pronoun
them in [ l ] functions as a constraint on relevance. It allows the writer to make sure that her
readers select the most effective contextual assumptions in their search for relevance at minimal
processing cost. As we have suggested, in this case such a contextual assumption is derived from
interpreting the semantic content of the whole preceding sentence in such a way that the second
fragment of [ l ] is interpreted as [l B]:
[ l B] Put them (the six washed and cored cooking apples) into a fireproof dish
IV. l . 1.b. Encapsulating discourse act

A second subtype of relevance of content that 1would like to propose takes place when, as well
as interpreting the semantic content ofthe whole of a preceding coherence unit, the reader needs
128 Ano I. ~bforeno
to interpret the discourse act performed by it in order to establish the propositional content of a
segment in the new sentence. In other words, the structure of the new sentence contains an
element that reclassifies a previous discourse segment in terms of its discourse function.
Examples of these signals of discourse acts may be: distinction, definition, difference,
comparison, or any other encapsulating device that refers to an act performed by some segment
of preceding discourse. Let us consider one example from the text:
(47) There will be logistical problems: retuming al1 scripts will mean 13.5m papers
whizzing through the postal system, for instance. Photocopying scripts sounds
horrendous even to a convinced "pro-retumer" like me. Proper scrutiny of the papers in
school will take time, possibly precious holiday time. (5 1) And if the big leamers here
are teachers, not pupils, should they be returned at all?
The answer is yes. 1 believe now, as I believed last year when 1 wrote one of the first
articles calling for this move towards long-overdue transparency and accountability, and as the
authorities hold inNew Zealand, that it is simply the right thing to do. The right thing overrides
logistical problems. (55) Pupil neglect of the papers is beside the point.
After reading these two paragraphs of the text, if we focus our attention on the noun
phrase the point in the last sentence. it is clear that it has an encapsulating function. The reader
will rapidly wonder which point, to remember that the point had been made in the preceding text
in the form of a rhetorical question "should they (the scripts) be returned at all?" which in its
context, where problems are being described, really implies what could be expressed by a
negative statement: they should not be returned at all. The answer given to this question by the
author, however, cancels this negative interpretation by providing what is presented as a
powerful argument: i.e. it is simply the right thing to do, and this overrides logistical problems.
Then, the last sentence, ( 5 9 , retrieves by means of the encapsulating devicepupil neglect
of the papers a previous assumption supplied by interpreting the text at coherence unit (26),
which was presented as a negative aspect (an irony) of the pilot scheme, i.e. that most of the
candidates didn't want them (i.e. the scripts). What the sentence in focus, ( 5 9 , does then is
discard the relevance of that assumption, (26), elaborated from (27) to (33), in relation to the
point made, i.e. the question posed in coherence unit (51). Thus in order for the reader to
establish the content of the noun phrase the point, s/he needs to interpret the discourse act
performed by coherence unit (5 1) as making a point.
Finally, encapsulating discourse acts can also co-occur with other deictic acts, such as
in this distinction. This reinforcement makes the encapsulation process easier to perceive.
IV. 1.2. Relevante of relationul jknction (Encapsulating logical act)

The other type of coherence arises when the information made available by one discourse
segment is used in establishing the relevance of the next (Blakemore, 1987: 1 12). In other words,
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 ( 1). 2003, pp. 1 11-165
The Rolr of Cohesivr Drvices as Trxtuol Constroints on Relevoncr 129
this type of coherence arises when, on trying to establish the relevance of the new coherence unit
as a whole (not simply one element in it -as in relevance of content), the reader needs to do
some extra inferential work to interpret the discourse function (i.e. an implicit import) of the
whole of a previous discourse unit in relation to the discourse function of the whole of the
current discourse unit. That is why 1refer to this type as relevance ofrelational function, because
the extra propositions inferred from the two related discourse segments are relational, i.e. depend
on each other, and they are perceived in discourse functional terrns. Other terrns used in the
literature to describe roughly the same kind of phenomena are the following: conjunctive
relations (Halliday & Hasan, 1976); semantic relations (Crombie, 1985); relational propositions
(Mann and Thompson, 1986); clause relations (Winter, 1986); intersentential relations (Hyde,
1990, 2002); and coherence relations (Sanders et al., 1993).
Examples of these relational propositions are: sequence, claim-support, argument-
conclusion, claim-contrast, reason-action, effect-cause, etc. Since it does not seem possible to
arrive at a consensus on a universal taxonomy, 1 will use the terrns that the subjects employed
intuitively to describe their interpretations. What most authors seem to agree on is that relational
propositions may be implicit or explicit. It is important to notice that it is when the relational
propositions are made explicit that the encapsulation is patent, serving as a powerful textual
constraint on relevance. Otherwise, the relevance of the new coherence unit can only count on
the reader's inferential capacity.
If we considered this phenomenon from Halliday and Hasan's (1976) view, then we
would be considering the encapsulating mechanism effected by conjunctive items, which include
expressions such as And, Yet, So and Then. However, as has been attested by Winter (1977),
Crombie (1985), Hyde (1990, 2002) and Moreno (1995, 1997, 1998b, in press), there are
alternative means of signalling relational propositions to the well-recognized conjuncts. These
alternative means stretch right across the spectrum of sentence structure, constituting central
elements such as nominal, verbal, adjectival and others items. Hyde (1990) provides a full
account of these means in his study of the explicit signalling of intersentential relations as they
occur in a corpus of editorials from The Guardian. Moreno (1995, 1995. 1998b) focuses on the
expression of different kinds of causal expressions in a corpus of research articles to offer
contrastive results between Spanish and English. Moreno (in press) identifies the full range of
possibilities for the expression of intersentential causal relations in a corpus of cause-effect
analytical essays.
It is precisely in most of these other alternative expressions where the mechanism of
encapsulation is perceived more clearly. The main reason is that these integrated signals usually
co-occur with other devices such as ellipsis, reference or lexical cohesion, which also
encapsulate, making the encapsulation stronger. Consider for example the metatextual expression
this is not to say identified in coherence unit (15) in the text analyzed, where the previous
relevant segment of text (12-14) is encapsulated by the reference item this, establishing
relevance of content. What is interesting to point out is that the subjects agreed that this
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES. vol. 3 ( l ) , 2003, pp. 1 1 1-165
130 Ana I. lifo~.eno
expression was also signalling a relation of inferred consequence derived from the previous
relevant discourse and that this relation was being cancelled by the negative word.
If we now consider Sinclair's (1993) analytical framework, relational propositions would
approximately correspond to what he terms logical acts. And this is the term 1will adopt to refer
to this phenomenon in order to avoid using more extraneous terminology. 1 say approximately
because, according to Sinclair (1993: 9), logical acts "show the use of the logical connectors and
associated mechanisms such as ellipsis", but in the present analysis ellipsis is not considered
under logical acts, but as a form of inferred connection, for the following reasons.
On the one hand, it is true that signals of relational propositions are also occasionally
associated with ellipsis such as in connectors like as a result (i.e. 'as a result * of a previously-
stated proposition'). This, in fact, has been well attested by Moreno (1997, 1998b, in press).
However, in a detailed analysis of the mechanisms of coherence involved in this expression, it
is possible to identify two distinctive sources of relevance which occur in combination: one
which arises when the second related fragment of text is interpreted as the result of what has
been stated in the previous relevant proposition, which is interpreted as the cause, thus inferring
a cause-effect relational proposition. The other source of coherence arises when the elliptical
element is recovered. It is then that the encapsulating mechanism is perceived. However, if we
had to analyze the encapsulating elements recovered, we would say that the type of relevance
they contribute to establishing has more to do with relevance of content. This type of relevance
would correspond more precisely to the coherence mechanisms Sinclair terms deictic acts.
On the other hand, ellipsis may also occur in expressions that do not signal a logical act.
Consider the following metatextual expression found in unit (26) in the text, the irony is. In this
case, the subjects interpreted that the expression was one form of evaluation of the proposition
previously stated in (6), where the pilot scheme had been first introduced and described. In fact,
the students were able to establish the connection with that previously stated proposition by
recovering the following textual material oj'the pilot scheme, which would be functioning as a
post-modifier in the NP: the irony oj'the pilot scheme. What is important to notice in this case
is that the encapsulation is not explicit, but implicit by means of ellipsis. It is only when the
elliptical material is recovered, not always an easy task with student subjects, that the
encapsulation is perceived. However, again, if we had to analyze the encapsulating elements
recovered, of the pilot scheme, we would say that the type of coherence they contribute to
establishing is relevance of content by means of reference (the) and lexical cohesion (repetition
ofpilot scheme). This is perhaps the major difference between the present analytical framework
and Sinclair's (1993).
Sinclair (1993) proves his hypothesis about the coherence mechanism temed
encapsulation, and concludes that the principal type of coherence is through encapsulation. It is
so well established that in cases where there is no explicit link between sentences the default
interpretation is encapsulation. Now, as Sinclair (1993) acknowledges, if encapsulation were an
absolute rule, and not just a default hypothesis, then the nature of text structure would be
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES, vol. 3 ( 1 ), 2003, pp. 1 1 1-165
The Role of Cohesive Devices as Textuul Construints on Relevunce 131
obvious. The current sentence would encapsulate the previous one, which in its turn had
encapsulated its predecessor, and so on back to the beginning of the text. The current sentence
would then be encapsulated in an act of reference in the next to come, and so on until the end of
the text.
As a model oftext structure, 1agree with Sinclair that it seems very attractive. It explains
how texts can be organized and how their dynamism may be created and fuelled. It provides the
basis for a powerful definition of coherence, and reduces cohesion to the identification of the act
of reference only. However, Sinclair (1993) introduces a variation to the default hypothesis,
which he considers as another major category of coherence: that is, prospection. This is an
altemative -but not so frequent- structure to that of retrospective encapsulation, which
Blakemore (1987) does not explicitly account for in her study.
IV.2. Prospective mechanisms (P)

In addition to encapsulating the preceding text, a sentence can make a prospection about the next
sentence, thus establishing a need for the next sentence to fulfil the prospection if coherence is
to be maintained. The sentence fulfilling the prospection does not encapsulate the prospecting
sentence. (Sinclair, 1993: 28)
So, prospection occurs where the phrasing of a sentence leads the reader to expect
something specific in the forthcoming text. Due to the precise nature of the type of relevance
established in cases of prospection, 1 will distinguish the following two types on the basis of
whether they are used to establish relevance of content or to establish relevance of relational
function.
IV. 2.1. Relevunce of'content

In this variation of coherence mechanism we can perceive a similar principle to the one we could
perceive in the corresponding type of encapsulation but in the opposite direction. Blakemore
(1 987,1992) does not acknowledge the phenomenon of prospection, because she focuses on the
relevance of the current utterance in relation to meaning generated by interpreting the preceding
discourse. However, from our relevance-based perspective it should be possible to say that one
type of prospection occurs where there are text elements in the current sentence whose
propositional content is likely to be affected by the interpretation of an upcoming text fragment
in the sense that its meaning will be fully determined.
Another way of looking at this is to say that prospection occurs when there is an element
in the current sentence that gives the reader advanced warning as to the way in which the
assumptions derived from interpreting the following segment of discourse will be relevant. Thus
in prospective acts of this type, an element of the current sentence serves as a powerful constraint
on the relevance of an upcoming fragment of text. It is also important to notice that this
phenomenon implies that the word or phrase to be elucidated in the upcoming text is presented
132 Ana l. Xloieno
as new to the context created in the course of interpretation. Within prospecting relevance of
content it is possible to distinguish at least two types of prospecting act.
IV.2.1 .a. Prospecting deictic act

The first type roughly corresponds to the phenomenon identified by Tadros (1985: 14) as
enumerution. It rests on the reader interpreting the full meaning of a word or phrase (e.g. a sub-
technical word such as udvantages, aspects, functions, which Tadros terms the enumerable), as
something to be elucidated in the following text. Tadros shows how the enumerable, or
prospecting signal, is usually preceded by some kind of numeral, whether exact, such as two, or
inexact, such as several that commits the writer to enumerate. However, as the text shows, the
enumerable does not need to be preceded by a numeral to create a prospection. It is sometimes
simply expressed in the plural. A clear example from the text is found in coherence unit (47):
there will be logisticalproblems, which by means of the lexical wordproblems (a superordinate),
followed by a colon, makes a prospection over a group of sentences (48-SO), which speci@ the
logistical problems prospected.
The distinguishing feature of this type of prospection from the relevance perspective rests
on the fact that interpreting the semantic content of a segment of upcoming discourse will help
to fully determine the meaning of the prospecting signal. This also has the effect of establishing
the relevance of the next fragment of discourse. What is more, for the prospection to be fulfilled
satisfactorily, the semantic interpretation derived from the following unit(s) needs to be
congruent with the general semantic meaning of the prospecting signal. For instance, relevance
is easily perceived at each of text units (48-50) when after interpreting their semantic content we
are able to abstract away and interpret each of the events described as problems.
1have temed this first typeprospecting deictic act in a general sense to include not only
this type of sub-technical lexical words, or superordinates, but also other prospecting signals
such as cataphoric reference items and question words, where the meaning of the prospecting
item is also elucidated by interpreting the semantic content of a relevant segment of upcoming
discourse. The question word, what, in coherence unit (2) in the text is a clear case. Punctuation
marks such as the question mark (2) or the colon (47) would also contribute to establishing a
prospection.
IV.2.1.b. Prospecting discourse act

Another common way in which this type of prospection may happen is when the current
sentence contains a signal, similar to what Tadros (1985) terms advance labelling, such as let
us define, whereby the "writer labels, and thereby commits himself to perform a discourse act"
(Tadros. 1985: 22). In this case, the writer is committed to performing an act of definition. In
other words, for the reader to fully determine the content of the element define in the current
coherence unit, s/he will need to go on reading the following relevant fragment of text and
interpret it as a definition. Thus, in this case too, the current sentence serves as a powerful
e
The Role ofCohesive Devices us Textuul Construints on Relrvunce 133
constraint on the relevance of the upcoming coherence unit, in that it will constrain its
interpretation to a given discourse function, a definition. In fact, if the reader is to perceive the
new coherent unit as relevant, slhe needs to be able to infer its discourse function as a definition.
It is this inferred discourse function that needs to be congruent with the general meaning of the
prospecting signal.
This type of prospection would embrace cases such as those accounted for by Sinclair
(1993). One exarnple is the introduction of quoted speech by means of attribution, as in his
messuge, the stutement or the exhortation (Sinclair, 1993: 13). Other possible signals of
prospected discourse acts may be the following: consider, discuss, compare, describe, examine,
mention and distinguish, as in a sentence like "It is important to distinguish between real and
nominal wages" (Tadros, 1985: 22) followed by other sentences elucidating this distinction. It
should be noted that the function of the following fragment of discourse is not part of a relational
proposition but is just an autonomous discourse act.
The only example of a prospecting discourse act found in the text under analysis is in
coherence unit (58):
(56)A few will be very interested indeed, (57) and that's enough. (58) * (It is) A bit like
voting, really: < [(59) lots of people don't care about that either, (60) but for those who
do, it's one of the markers of a civilized world.]
After reading (58), it seems as if the writer is committed to performing an act of

comparison. It is true that a comparison is made in this clause by means of the comparative
preposition, like, between the situation encapsulated by elliptical material such as it is and voting.
In this sense, like is encapsulating, because the reference of the comparison is found in previous
text. However, the comparison is not fully determined in the clause where like occurs, since the
reader does not know in what way the two members of the comparison are similar. To satisfy
this, the reader will need to go on reading. In this sense, the comparative preposition is
prospecting a discourse act of comparison. Reinforcing this prospection is the colon, which
indicates that the fulfilment of the prospection will follow immediately.
IV. 2.2. Relevance of relational function (Prospecting logical act)

Another variation of prospection that 1would like to propose serves to help readers perceive the
relevance of a new coherence unit by advancing the relational discourse function that will be
established between the next fragment of discourse and, either the current sentence, or a previous
fragment of discourse. In other words, prospecting logical acts also play an important role as
textual constraints on relevance in the sense that by interpreting the current coherent unit, the
reader is able to predict the relevance of the next in terms of its relational discourse function.
That is, in this type of prospection some proposition (or pragmatic import) derived from
interpreting the current segment of discourse is used in establishing the relevance of the
134 Ana I. Moreno
following segment of discourse by virtue of its discourse function in relation to the discourse
function inferred from the current sentence o r a previous one.
An interesting example from the text is in (28), the reasons are obvious, where the
prospecting signal is the plural noun reasons. It is true that this case might also be analyzed as
a case of enumeration (i.e. as a prospecting deictic act), in the sense that the content of the word
reasons will be elucidated in the following text. That is, the reader will need to go on reading
the following fragment of discourse to find the reasons enumerated. However, it also seems quite
clear that interpreting the coherence unit in which the signal appears leads the reader to predict
the relevance of the upcoming unit(s) in discourse functional terms. In the present case, the
reader is led to interpret the following fragment of text as the reasons for the previous relevant
discourse, which is then interpreted as the fact or action that will be justified. Therefore the
reader infers a relational proposition of fact-explanation or action-reason, which helps h i d h e r
establish relevance of relational discourse function for the forthcoming piece of discourse.
In this particular case, what can also be observed is that the fragment of discourse from
which one member of the relational proposition is inferred (Le. the action to be justified) is in
a previous independent sentence. Thus in order to perceive the relevance of the current sentence
the reader needs to recover the previously stated proposition which has been omitted in the
current sentence. On recovering elliptical discourse material such as for the previously stated
action, a mechanism of encapsulation is clearly perceived as establishing relevance of content.
So, we could say that this coherence unit is perceived as coherent by virtue of dependent
relevance of relational function both retrospectively and prospectively, and by virtue of
relevance of content prospectively, by means of the signal reasons.
It is in cases like these where it is possible to appreciate the role of prospecting cohesive
devices as another type of constraint on the relevance of the upcoming piece of discourse.
Blakemore (1 987, 1992), however, does not account for this kind of relevance. Sinclair (1 993)
and Tadros (1985), for instance, do account for some of these prospecting signals such as
examples, implications in their role as what 1have called constraints on relevance of content, but
they do not clearly account for their role as constraints on relevance of relational function. Hyde
(1 990), on his part, gives full account of this role in his discussion of cataphoric and anaphoric-
cum-cataphoric intersentential relations signals which occurred in editorials from The Guardian
throughout the whole spectrum of relational propositions: causal, adversative, additive and
temporal.
IV.3. Units fulfilling or satisfying a prospection (S)

As Sinclair (1993) puts it, the prospective acts relevant to a sentence are made in the previous
sentence. The act of prospection means that the interactive force of a sentence extends to the end
of the sentence following. If we now look at this phenomenon from the relevance-based
framework, we will also need to account for the relevance of the unit following a prospective act.
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES, vol. 3 ( l ) , 2003, pp. 11 1-165
The Role qf Cohesive Devices us Textuul Constraints on Reltvunce 135
1 would like to suggest that the relevance of that upcoming unit, which becomes the current
sentence in the process of interpretation, is perceived if it satisfies the prospection made in the
previous text. The prospection may be fulfilled in two ways: a) if the current sentence provides
information from which to derive assumptions (in terms of semantic content or discourse act)
that may be used to determine fully the content of a part of the propositional content of the
coherence unit where the prospection was created, andlor b) if it provides information from
which to derive a relational discourse function congruent with the relational proposition
prospected in the preceding discourse. Failing this, the reader may find the discourse either
unsatisfactory, incomplete or illogical.
In the case under analysis, every sentence in the rest of the paragraph following the
reasons are obvious is relevant in this sense. Al1 these sentences together are then said to fulfil
the prospection. And their status in the text structure will be that of fulfilling the prospection. Let
us imagine, for instance. that instead of coherence unit (28) being continued in the way it was,
it had been followed by the following coherence unit: r y o u had two older brothers, you really
don 't care ubout thepapers. Even though the relevance of this upcoming segment of discourse,
now the current text of the moment, may have been prospected by the previous coherent unit the
reasons are obvious, the reader will very probably find it difficult to derive any assumption from
the information contained in it that allows himlher to infer that the information is functioning
as one of the prospected reasons. Therefore, the reader is bound to find the text illogical or
irrational at this (invented) point. This is why 1 would like to stress the role of the fulfilment of
a prospection as a powerful though less frequent coherence mechanism.
In this section 1 have introduced the main criteria used to analyze explicit coherence
mechanisms and have discussed their role as textual constraints on relevance. It should be
emphasized that this method of analysis was applied only to those text features identified by the
subjects, whose contribution to establishing the relevance of each new sentence was discussed
open-class. In summary, these text features were classified either as encapsulating (deictic act,
discourse act, logical act), prospecting (deictic act, discourse act, logical act), combining both
mechanisms, or fulfilling a prospection.
IV.4. Inferred encapsulation, or qualified assignments (1)

In cases where there were no clear explicit signals of the coherence mechanism, that is, in cases
of implicit connections, the subjects were asked to make them explicit. These were the cases that
roughly correspond to what Sinclair (1993:20) terms qualified assignments. He also suggests
that, as a general rulé in interpretation, in the absence of a clear indication we reverse the
argument and ask what is the kind of relationship that, using al1 the powers of inference
available, one would assume in that case. As a method for gathering data, that is exactly what
the subjects were asked to do. Then the group tried to arrive at a consensus about the most
reasonable interpretation in relation to the groups' standard of coherence, which may not be the
136 Ana l. ,bfo~ior.eno
same in other discourse communities. This recovered textual material typically signals logical
acts and deictic acts of an elliptical type.
It is important to stress that the category of ellipsis has been treated in the present study
either as a case of inferred point-to-point cohesion or inferred encapsulation since one of the
characteristics of this cohesive tie is precisely that there is no text signal indicating the tie but
a structural slot that needs to be recovered for relevance to be established.
In some cases the structural slot is obligatory from a syntactic viewpoint as in (23).
(23) * (Anyway) Even if he didn't (*loom over one 'S shoulder, checking, commenting,
re-marking ifnecessary), the fear that he would (*loom over ones shoulder, checking,
commenting, re-marking ifnecessary) was a great deterrent to misdemeanour.
In this case the elliptical text segment refers to a part of the wording used in the previous
coherence unit (22): the predication in the clause. This would be a case of point-to-point
cohesion. In other types of ellipsis, the structural slot is optional, as in (7).
( 7 ) The Qualifications and Curriculum Authority has carried out an interim evaluation
* (ofthe pilot scheme).
In (7),the elliptical encapsulating item did not simply refer to the part of the wording in
the preceding coherence unit, ( 6 ) where the pilot scheme is first mentioned but to the whole
semantic content of (6), where the pilot scheme is described in detail. Once recovered, the
encapsulating devices were analyzed and classified as any other explicit encapsulating acts. If
one looks at these and other cases of ellipsis closely, two types of relevance seem to arise once
the elliptical material is recovered.
ZV.4.1. Relevance ofcontent (Znferred encapsulating deictic act)

One is relevance of content, as illustrated by (7),where the interpretation of one segment in the
current sentence (excluding the linking word oj), the pilot scheme, is affected by the
interpretation of another segment of previous discourse, i.e. the whole of coherence unit ( 6 )
where the pilot scheme is described. It is this phenomenon that can be considered as really
textual in nature because it involves encapsulation rather than point-to-point cohesion.
IV. 4.2. Relevance ofwording (Inferredpoint-to-point wording act)

A second subtype of relevance that 1 would like to propose, drawing on Halliday and Hasan
(1976), takes place when, rather than recovering the semantic content of the whole preceding
coherence unit, the reader needs to recover (a part of) the wording used in it in order to establish
the content of the elliptical segment in the new sentence, as in the two instances of ellipsis in
(23)above. As we shall see, this type of point-to-point cohesion is usually accompanied by other
The Role qf Cohesive Devices as Textuul Consrruints un Relevunce 137
types of cohesion, such as he in (23), which are able to encapsulate. It is worth noting that this
type of relevance would also apply to cases of substitution, although in these cases the relation
is made explicit by a word such as one and do. These cases also seem to reflect point-to-point
cohesion rather than true encapsulation, as will be shown.
IV.5. Type of cohesive tie

Once al1 the sentences in the text had been classified according to the type of coherence
mechanism that helped the readers to perceive their relevance, the study sought to determine
which type(s) of cohesive tie waslwere involved in each case. One problem at this stage was to
decide which taxonomy of cohesive devices to use to classify the different coherence
mechanisms found. It was eventually decided to use Halliday and Hasan's (1976) classification
of cohesive devices for the simple reason that it is still the most comprehensive and widely
known account of cohesive devices. Therefore, using their terminological framework would
make it easier for researchers to establish comparisons between results obtained applying
different but related models.
Thus, the textual features identified by the group as contributing to their perception of
coherence were further classified, wherever possible, under the different categories identified
by Halliday and Hasan (1976): fexicafcohesion (Le. repetition, synonym, superordinate, general
word, related word); substitution (nominal, verbal, clausal) reference (personal, demonstrative,
comparative). Logical acts were classified on a first leve1 according to the four types of
conjunction (additive, causal. adversative, temporal). In cases where the identified features did
not fit any of these categories, further categories were added. Within conjunctive relations,
further subcategories were specified but, as has been mentioned, the terminology used to name
each logical relation in some cases had more to do with the subjects' interpretation of the
relational discourse functions inferred than with the categories used by any of the existing
studies in the literature to avoid losing the shades of relational discourse meaning perceived.
Finally, as justified above, the category of effipsiswas treated as a case of either inferred
encapsulation or point-to-point cohesion.
IV.6. A text-as-product view of cohesive devices

The last stage in this research was to analyze the cohesive ties contained in the text when looked
at as a product. By this 1 mean using a method whereby the researcher approaches the whole text
as a finished product in an attempt to identi@ al1 kinds of cohesive ties that play a role in
establishing connections between a text fragment and another one across sentence boundaries.
Due to obvious space limitations, it was beyond the scope of this paper to classi@ al1 the
different resources found. Therefore, these are only marked in the text. However, let me point
out that, in order to make future comparisons possible, the identification of these devices was
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 (l), 2003, pp. 1 1 1-165
138 Ana l. ,Vfot,eno
carried out following Halliday and Hasan's (1976) classification of cohesive ties as far as
possible: lexical cohesion (repetition, synonyms, opposites, related words), substitution and
ellipsis. reference Cjust endophoric) and conjunction.
V. RESULTS
The results obtained from the discourse-as-process view of textual constraints on relevance are
summarized in Table 2.
Table 2: Textual constraints on the relevance of each coherence unit in the text
Type ond subgpe
N" Esplicit signal(s)
of colrerence Type of colresive tie Relates to
* =Inferred signal(s)
mecltartism
1
P: Deictic Reference: Question word

Punctuation: Question mark
2) and E: Logical Additive: Positive

3) the water E: Deictic
1 Lexical repetition 1 1
b T he water 1 E: Deictic / Phrasal: Demonstrative reference + 1 1-4
1 Lexical repetition
1 E: Deictic 1 Phrasal: Demonstrative reference + 5
1 1 scripts ... to candidates 1 1 Lexical repetition with word class 1 ¡
change + Synonym + Repetition
1) * (of the pilot 1: Deictic 6
scheme)
2) an evaluation P: Deictic Lexical: Superordinate 9-1 1
S 1) it - E: Deictic Reference: Personal 6
9 1) S 7
2) * (about the pilot I Elliptical: Post-modifier 6
1 scheme)
10 1) 1S 7
2) * (In other 1: Logical Additive: Expository:
words)
1)
2) It E: Deictic Reference: Personal
1) * (ln fact) 1: Logical Causal: Claim-support
2) * (in the pilot 1: Deictic Elliptical: Post-modifier
scheme)
1) * (because) 1: Logical Causal: Claim-reason
1) * (In other words) 1. Log~cal Causal. lnferred consequence
1) This is not to say E: Deictic Reference: Demonstrative
that E: Logical Causal: (Cancellation of) inferred
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 (l), 2003, pp. 11 1-165
The Role of Cohesive Devices os Texti~alConstroints on Relevonce 139
m a thing T a c T ~ h r a s a lComparative
: reference + 1 5 7 (
Additive: Expository: Paraphrase

Adversative: Proper 19
Causal: Claim-reason .. 20
2 2 1) ~t least 1 E: Logical 1 Adversative: Corrective 1
1 1 2) that + he did ( E: Deictic 1 Clausal: Same meaning: 1 1
Demonstrative reference +
Personal reference + Leaicnl general verb
1) * (Anyway) 1: Logical Adversative: Dismissive 21-22
2) he E: Deictic Clausal: Personal reference + 21-22
3) did * (loom over
one's shoulder ...)
4) he
5) would * (loom over
1 (PP): Wording
E: Deictic
I (PP): Wording
Predicator ellipsis
1 Clausal: Personal referente +

Predicator ellipsis
1 Predication
21-22
Predication
1
Adversative: Contrastive
E: Deictic Reference: Comparative
1 1 3) * (than the chief 1: Deictic ) Elliptical: Comparative clause 1 1
examiner looming
over ...)
1) * (Retuming the 1: Deictic Elliptical: Subject + Operator
marked scripts... is)
1
--
,
2) * (of the pilot 1: Deictic Elliptical: Post-modifier
scheme)
3) * (in spite of...) E: Logical Adversative: Proper
having been offered E: Deictic Clausal: lnferred meaning
l
matter of 1: Logical Causal: Claim-support

fac t
b8 1) $he reasons 1 P: Deictic Demonstrative
Lexical superordinate
EP: Logical Causal: Action-reason
2)* (why most of the 1: Deictic Elliptical: Clause
1 1 students did not want 1 1 1 1
2) and E: Logical Additive: Positive 29

E: Deictic
- - Reference: Demonstrative
1: Logical Adversative: Contrastive
/
/ 2) lnterest in the
a ers is enerated 1 E: Deictic
1 Clausal: Opposite meaning
(situam
l I
32
33 1 2) and * (and = but)

3) then * (then = that)
1; * (because)
2 If ppp
E: Logical
E: Deictic
1: Logical
Adversative: Proper
Reference: Demonstrative
Causal: Claim-support
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES,vol.3(1),2003,pp. 111-165

3) you partied al1 year, ' E: Deictic Clausal: Opposite meaning (instance)
or had a personal
w-
35
36
-p.
crisis
1) * (this is)
1) * (because)
E: Logical
1: Deictic
1: Logical
Additive: Positive
Elliptical: Subiect + Operator
26-33
34
35
37 1) Actually, E: Logical Additive: Avowal 36
2) this -. E: Deictic Reference: Demonstrative
38 1) * (because) 1: Logical Causal: Claim-reason . 37
39 1) * (ln fact) 1: Logical Causal: Claim-support 38
40 1) that requirement E: Deictic Phrasal: Demonstrative reference + 39
Lexical superordinate
41 1) lf * (lf = while) E: Logical Adversative: Proper 26-33
2) candidates didn't
care about the scripts E: Deictic Clausal: Similar meaning
42 1) * (As a matter of 1: Logical Causal: Claim-support 41
fact)
43 1) Of course * (Of 1: ~ e i c t i c Elliptical: Subject + Operator
44
45
course = this is
natural)
1) * (because)
1) * (However)
2) Better than
3) knowing what they
got * (which is the
1: Logical
1: Logical
E: Deictic
E: Deictic
Adversative: Proper
Reference: Comparative
p-
Clausal: lnferred meaning
44
best learning tool a

--
1: Logical Causal: lnferred consequence 45

E: Deictic Reference: Demonstrative
P: Deictic Lexical: Superordinate + 48-50
Punctuation: Colon 1
2) * (with the process 1: Deictic Elliptical: Post-modifier 6
1 1 of returning the scripts 1 1 l 1
1 to candidates)
48 1 1) 1S
?
E: Logical Additive: Exem lificato
S
S
E: Logical Additive: Positive
2) if E: Logical Causal: Conditional
3) the big learners are E: Deictic Clausal: lnferred meaning
1 teachers, not pupils 1
52 1 ) * (However) 1 1: Logical
to this question) 1 1: Discourse Elliptical: Post-modifier
3) * (they should be 1: Deictic Elliptical: Clause
retumed)
53 1) * (because) 1: Logical Causal: Claim-reason 52
54 1) * (And) 1: Logical Additive: Positive 5 3 7
2) The right thing E: Deictic Phrasal: Demonstrative reference +
Lexical repetition
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. / E S , vol. 3 (1), 2003, pp. 1 11-165
The Role o f Cohesive Devices us Texluul Conslruinls on Relevunce 141
;) -
roblems 1 E: Deictic 1 Lexical: $perordinate
55 1 1) * (On the other ) 1: Logical 1 Additive: Positive
hand 1 Also)
2) pupil neglect of the E: Deictic Phrasal: Lexical synonym + Synonym
papers with word class change + Synonym
Phrasal: Demonstrative reference +
3) the point E: Deictic + Lexical superordinate
Discourse
1) * (because) 1: Logical Causal: Claim-support -
1) and E: Logical
2) that
1) * (lt is)
2) like
E: Deictic
1 : Deictic
E: Deictic
t )
Referente: Demonstrative
Elliptical: Subject + Operator
Reference: Comparative
3) like P: Discourse 1 Reference: Comparative
4) : P: Deictic Punctuation: Colon
1) S
2) that PP: Deictic Reference: Demonstrative
3) either PP: Deictic Reference: Comparative adverb
1) S
2) but E: Logical Adversative: Proper
3) those * (people) PP + I (PP): Deictic Demonstrative reference + Nominal
I (PP): Wording ellipsis
4) do * (care) Ellipsis or Substitution: Verbal Predicate in
PP: Deictic (59)
Reference: Personal Noun in ( 5 8 )
The first column indicates the number of the coherence unit being analyzed. The second column
indicates which signal(s) were reported to make the connection between the current unit and its
co-text explicit. In cases where the connection is implicit, the results show in brackets and
preceded by an asterisk which signal(s) were provided by the subjects to make the connection
explicit. The third column first specifies the type of coherence mechanism perceived (E =
encapsulation, 1 = inferred connection, PP = point-to-point cohesion, P = prospection. EP =
encapsulation-cum-prospection, S = satisfaction or fulfilment of prospection). What comes afier
the colon specifies the subtype of mechanism involved on the basis of the type of relevance
perceived: (Deictic / Discourse = relevance of content; Logical = relevance of relational
function, Wording = relevance of wording). The fourth column shows the type of cohesive tie
identified. The fifih column indicates in relation to which part of the co-text the connection is
established. This may be specified by: the number(s) of the encapsulated or prospected
coherence unit(s); the number of the coherence unit where the prospection was created; or, when
the related element is smaller than a clause, its grammatical function in the clause.
These text features can be observed in their full context in the segmented text found in
the appendix. They appear in bold type. The underlined items can be considered as the other text-
as-product cohesive mechanisms. Thus, although they may contribute to the perception of
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 ( 1 ), 2003, pp. 1 1 1-165
142 Ana I. A401,eno
superficial cohesion in the text, they are not essential to perceiving its coherence and text
structure at least by these subjects. This does not of course mean that the cohesive signals
identified in this paper as textual constraints on relevance from a discourse-as-process
perspective would not also be identified as cohesive signals from a text-as-product perspective.
It is most probable that a text-as-product view would have identified almost al1 these cases too,
except for most of the inferred items.
Table 3 offers a summary of the coherence mechanisms affecting each coherence unit in
the text that account for the coherence of this text as perceived by the subjects.
Tuble 3: Major coherence mechanisms affecting each coherence unit

Coherence mechanism N % Number of coherence unit
12. 13. 14. 17. 18, 19.21.25.27.35.36.38.39.
19 32
Iiifcrrcd ciicapsiilatioii 42.43.44. 52, 53. 56
tulfilniciit o f prospcctioii + Iiifcrrcd E. 2 34 9. 10

t'ulfilinciit o f prospcctioii + Point-10-poiiit colicsioii 1 1.7 59
t'iilfiliiiciit of prospcclion + Eiicapsiilatioii + Iiifcrrcd point-to-poiiit
1 1.7 60
coIicsioii + Point-10-poilit colics~o~i
VI. DISCUSSION
The first finding worth noting is the fact that the subjects were able to perceive a connection of
some kind between each new unit of coherence and its co-text. In cases where they could not
identi@ an explicit textual item marking the connection, the subjects were able to make the
relation explicit by inserting some kind of textual element. As 88% of the subjects reported in
a post-test questionnaire, having to make this interpretation task explicit was very useful in
helping them to perceive the coherence of the text at each juncture. Thus, it has been possible
to assign al1 coherence units to one of the coherence mechanisms listed in table 3, some of which
are combinations. In this sense, the framework proposed can be considered as satisfactory in
accounting for the coherence pattem of this text. Let us now comment on the most salient
findings, referring especially to those cases that have been difficult to classi@.
VI.l. Inferred encapsulation

In this particular text, the percentage of inferred encapsulations that account for coherence on
their own is relatively high (32.2%). So, if it were possible to establish a correlation between the
degree of implicitness of coherence relations in a text and the leve1 of difficulty in perceiving
Thr Rolr of Cohrsivr Drvicrs as Trxtuol Constroints on Relevancr 143
the coherence of its discourse, the difficulty in perceiving the coherence of this text might be said
to be in principle relatively high. And so it was, as the majority (80%) of subjects acknowledged.
However, after the group discussion it was possible to arrive at a consensus on most cases
without much difficulty. This points to the existence of a standard of coherence shared by this
discourse community that goes beyond the presence or absence of explicit signals.
VI. l .l . Overlay revisited (Inferred additive expository logical act by means of a paraphrase)
A number of cases that deserve special attention are what Sinclair (1993) could have termed
overluy within his methodological framework. In overlay "there is no obvious act of reference
in a sentence with respect to the one before it, and yet the two appear to be closely connected
-in fact, they are often almost paraphrases of each other. In such cases the new sentence takes
the place of the old" (Sinclair, 1993: 17). For instance, coherence unit (1 0) was interpreted as
a metaphorical paraphrase of (9), and (19) was interpreted as a paraphrase of (18). Let us
consider the relation between (19) and (1 8).
(17) 1examined for years: (18) the most gruelling job in the world, requiring painstaking
effort and concentration to sustain standards justly for 300 scripts in three weeks. (19)
Conscientious marking is a killer.
Both coherence units, (1 8) and (19), refer to the process of exumininglmarking done in
a conscientiouslpainstakingly way to say that this task was the most gruellingjobla killer. It is
interesting to notice how there is no single element in sentence (19) that may be said to
encapsulate the whole of the previous sentence on its own, leaving the rest of the sentence to
develop n~eaningfurther. In this case, the whole sentence paraphrases the previous one without
making the discourse advance at this point. The second sentence is simply another way of
expressing the same idea perhaps to strengthen, more or less forcefully, the assumption derived
from coherence unit (1 S), as could be explained from the relevance framework. In fact, when the
subjects analyzed this relationship, they unanimously agreed that they could perceive an
expository additive relation, which they could make explicit by means of a conjunct such as in
other words.
Now, if we agree that an additive relation of this kind is a logical relation, as many
accounts of cohesion do (Halliday & Hasan, 1976), then we could say that the relation between
these two coherence units is just another case of an inferred logical act. If we use a phrase like
in other words, or an equivalent expression such as let me express tlzis in other words, then we
may be able to see this as a case of implicit encapsulation, this being the explicit encapsulating
item. Thus, this interpretation saves us from having to interpret this phenomenon as an exception
(Sinclair, 1993: 16), but rather as a qualified statement. This way Sinclair's model may be made
even more powerful in accounting for discourse coherence than it looks, because it would reduce
the number of exceptions.
144 Ano l. Mot.eno
A smaller number of inferred encapsulations have been perceived in combination with

other types of coherence mechanism such as: encapsulation (13.6%); encapsulation and inferred
point-to-point cohesion (1.7%); encapsulation and prospection (3.4%); prospection (3.4%); and
fulfilment of prospection (3.4%). That is, although there was enough evidence in the text to
account for the relevance of some coherence units by virtue of explicit mechanisms, the subjects
were still able to infer other connections. This might be interpreted as the subjects' need to
reinforce relevance in order to make better sense of the text at that point in the interpretation of
the discourse.
VI.2. Encapsulation
As far as explicit encapsulation is concerned, it was easy to classi@ al1 acts according to the type
of relevance they contributed to establishing (deictic, discourse, or logical). Furthermore, most
of the cases of encapsulation could be classified under any of the categories of cohesive tie
established apriori by the method described above. However, there were some problems with
certain deictic devices that led me to create new categories that supplemented the well-
recognized categories of reference and lexical cohesion. It should be noted that on some
occasions these two categories occur in combination in the form of a nominal phrase (cf.
Examples (S), (16), (54) and (55)).
VI. 2.1. Encupsuluting clausul deictic uct

The first new category is what 1have temed clausal deictic act because the text segment of the
current sentence that is affected by the interpretation of the semantic content of a previous
fragment of text is not realized lexico-grammatically by a reference item andlor a lexical item
but by a whole clause. Various subtypes of encapsulating clauses have been noted according to
the type of meaning they convey in relation to previous discourse.
VI.2.1.a. Similar meaning (similar meaning clause)

This type occurs when a clause in the current sentence expreses meaning similar to the meaning
retrieved from a previous text segment in order to develop it further. One example can be
observed in the relationship between the following two sentences, which both serve to open two
different paragraphs in the text.
(26) The irony is, of course, that having been offered their scripts, most of the candidates
didn't want them.
(4 1 ) If candidates didn 't care about the scripts, 71 % of staff cared a great deal;
In (41) the propositional content expressed by the seemingly conditional clause

candidates didn 't care about the scripts can be considered as a clausal rephrase of the semantic
content expressed in (26) -excluding other metatextual elements both textual and interpersonal,
The Role o f Cohesive Devices as Textual Constraints on Relevance 145
such as: the irony is, of course, and having been offered their scripts. However, it is important
to highlight the distinction between overlay, or, in my view, inferred additive expository logical
act, and the present type of deictic act. The former rests on the interpretation of a relational
proposition between two related discourse segments, and the latter rests on the interpretation of
the semantic content of a previous segment of text. In the former it is the whole current
coherence unit that may be said to paraphrase the whole of the previous coherence unit. By
contrast. in the latter only one element in the current coherence unit - a clause- encapsulates
a previously stated proposition. The rest ofthe coherence unit is used to develop mutually shared
meaning further. In other words, while in the logical act the whole sentence paraphrases the
previous one without making the discourse advance at that point in terms of propositional
meaning, in the deictic act there are other elements in the same coherence unit that convey new
assumptions.
It is true that lexical point-to-point cohesion is involved in the clausal deictic act under
consideration, such as repetition by means of candidates and scripts (notice that them in (26)
refers back to scripts in the same sentence), and synonymy by means of care (referring back to
want). However, what helped the subjects to perceive the relevance of sentence (41) was the
appreciation that a whole idea previously stated in (26) was taken up again by (41) and was
reused as part of the propositional content of (41). In fact, the subjects observed that this whole
idea was used again to retrieve the relevant context in which introducing information such as
71% of staffcared a great deal was perceived as something totally unexpected. It is interesting
to notice how, in spite of there being a typical marker of a conditional relation if; the subjects
interpreted this relation as an adversative proper one and they even provided an altemative signal
such as while. Since the subordinate clause encapsulates the whole of a previously stated
proposition, the adversative proper relation in (41) is analyzed as intersentential rather than
intrasentential, therefore contributing to the subjects' perception of textual coherence.
VI.2.1 .b. Inferred meaning (Inferred meaning clause)

An interesting variation of clausal deictic act occurs when a clause in the current coherent unit
is interpreted as conveying meaning that may be inferred from a previous segment, and that
meaning is further developed in the current sentence. This can be observed in the relationship
between (51) and the whole of a previous paragraph (41-46).
(41-46) If candidates didn't care about the scripts, 7 1% of staff cared a great deal: 82%
agreed that access to the scripts would help with teaching the syllabus in the coming year.
Well of course. Knowing exactly where the last candidates got it wrong is the best
learning tool a teacher can have to improve performance next year. Better than knowing
what they got is knowing why they got it. If any government wants to conjure up massive
whole school improvement, this is the magic wand.
(S 1) And ifthe big learners here are teachers, notpupils, should they be retumed at all?
146 Ana I. kforeno
It is worth noting how coherence unit (5 1) encapsulates the meaning created by paragraph
(4 1-46). However, in this case instead of rephrasing the semantic meaning conveyed by that
previous fragment of discourse in a straightforward way, the author takes up the meaning
conveyed by the whole of that paragraph and expresses it in the form of an inference, or
interpretation, that may be derived from it on the following assumption: the fact that those who
have leamed from the pilot scheme are those to whom it was not in principle addressed is
negative. However, this assumption is not made explicit in the text. It needs to be brought to bear
by the reader, and it is the writer who manipulates the retrieval of that assumption by means of
persuasion. The persuasive strategy here was easy to perceive.
Adding this inferential assumption to a sequence of negative points (stated from (48) to
(50), cf. the segmented text in the appendix) makes it clear to the reader that the fact that the big
learners here are teachers, notpupils should also be interpreted as a problem or negative point.
That is, by using an inferential rephrase in the appropriate place the author has managed to make
accesible to the reader the most relevant assumptions the reader needs to bring to bear in order
to interpret that assumption as another negative point which makes it easier for the reader to
accept the authors's new point in coherence unit (5 1): should the scripts be returned at all?
VI.2.1 .c. Opposite meaning (opposite meaning clause)

Another variation of clausal deictic act occurs when a clause in the current coherence unit
expresses the opposite meaning to the meaning conveyed in the whole of a previous fragment
of discourse in order to develop it further. This has taken two forms in the text: A) by expressing
the opposite situation; and B) by expressing an opposite instance.
A) Opposite situation: If we look at the connection between (31) and previous discourse,
it will be easy to realize that a clause in (3 l), i.e. interest in thepapers is generated, expresses
the opposite situation to that expressed in (28). In other words, this clause serves to introduce
the reason(s) for the opposite situation expressed in (28), i.e. the reason(s) why students were
interested in viewing the papers. The rest of the clause, by doing badly, then develops that
reason.
(28) The reasons * (why most of the students did not want to view the scripts) are
obvious: if you did well, you really don't care about the papers -and that goes for doing
well unexpectedly, as well as having the satisfactionof achievingjust what you expected.
(3 1) Interest in thepapers is generated by doing badly, (32) and then only if it surprises
you.
B) Opposite instance: A clear example of the second variation can be seen in the
relationship between (33) and (32):
The Role o f Cohesive Devices us Textuul Construints on Relworice 147
(3 1) Interest in the papers is generated by doing badly, (32) and then only if it surprises
you. (33) Ifyoupartied al1 year, or had apersonal crisis, then you will have done badly
but you won't need to see the papers to see why.
In (33) we observe that the propositional content expressed by you partied al1 year, or
hud apersonal crisis conveys two instances which contradict or oppose the claim made in (32)
doing badly surprises you. In other words. the two instances provided by the conditional clause
compound express events that would not allow anybody to be surprised at doing badly. It is
possible to state that clause compound (33) encapsulates previous meaning because it does not
really add new propositional material. It just presents it on a more specific level by means of an
instance, or instances. In addition, the two events change the polarity of the proposition in
relation to which they are relevant. The encapsulation could, in fact, be paraphrased in more
general terms as: ifit does not surprise you or otherwise. This is what allows us to analyze the
conditional relation as one of reversed polarity (cf. Halliday & Hasan, 1976: 259). This is also
why the conditional relation can be considered as intersentential rather than intrasentential:
because it really serves to link the propositional content of the claim made in the matrix clause
of (33) to the propositional content of (32) by means of a reversed polarity conditional relation,
therefore crossing over sentence boundaries.
This relationship of opposite meaning would be parallel to the relationship of lexical
opposition identified in most accounts of cohesion but there would be an important difference.
While lexical opposition normally operates at the lexical level to establish a point-to-point
relation between two words or phrases, in which the meaning of one of the members of the pair
conveys the opposite meaning to the other member, opposite meaning clausal cohesion would
operate at the propositional level. That is, the semantic relation would be similar but it would be
between two propositions, each of which would be on opposite sides of the antonymic scale,
whether on the same or different degree of generality.
Now, despite there being enough evidence of the process of encapsulation between (33)
and (32), the predominant relationship that the subjects perceived between these two coherence
units was one inferred logical relation of support-claim. As they reported, unit (33) was
interpreted as offering support to the claim expressed by (32), which in its tum encapsulates
(3 1).
VI.3. Textual versus Point-to-point cohesion

Let us now go back to the open debate about the role of point-to-point cohesion in helping to
establish the relevance of text fragments by looking at the data provided by this text. There are
some cases identified by the subjects as really crucial that might be considered as cases of point-
to-point cohesion. For instance, it might be said that the phrase the water in ( 4 ) refers back to
148 Ana L Ifot.eno
the NP the water in ( 1 ) . However, the present model has reconsidered some of these cases in the
following way.
By virtue of relevance theory, the reader will bring to bear the most cost-effective
assumptions on the interpretation of the current sentence. In written language, these are most
likely to be derived from the immediately preceding text. As we have also said, the reader does
not usually retain the linguistic properties of previous text but the meaning created by it in the
form of assumptions. By virtue of the cooperative principle, the writer can rely on the fact that
the reader will retain the meaning conveyed so far by her discourse to some extent. So much so
that it will not be necessary to remind the reader of al1 the assumptions already shared at every
point in the text. From this perspective, it may be said that phrases such as the water in (4) and
(5) also function as textual constraints on relevance. They act as pointers that help readers to
retrieve whatever previous assumptions are needed for the interpretation of the new sentence,
not just the meaning created by the repeated phrase in previous text.
In fact, when the subjects of this study were asked about the assumptions they had used
in interpreting sentence ( 9 , for instance, they acknowledged that the repetition of the phrase the
water did not make them think ofwater in general but the water to which the author had referred
metaphorically as the water in which we dipped a toe in 1999, where the sharks didn 't bite and
which was not freezing. Thus, the new sentence was seen as a sentence that reuses
4ncapsulates- the ideas created in (1 -4), to develop them further by means of identification,
in such way that it helps to resolve the metaphor, a device that may have been used to attract the
reader's attention. In other words, what the readers of this study seem to have interpreted on
reading ( 5 ) is that:
( 5 ) The water [in which we dipped a toe in 1999, where the sharks didn't bite and which
was not freezing] was the great scary ocean of returning examination papers to
candidates.
There are other cases where the encapsulation does not seem to affect the whole of the
previous relevant piece of text but only a smaller part. For instance, the encapsulation effected
by the combination of deictic acts observed in (6) by means of the...return o$ .. scripts... to
candidates seems to operate only over the meaning created by returning examinationpapers to
candidates in (5).
(5) The water was the great scary ocean of returning examination papers to candidates.
(6) This year saw the pilot scheme, with three different models for GCSE and at A level,
for the copying and return of al1 scripts in 10 syllabuses, allowing centres to decide how
to release the copied scripts to candidates.
The Role of Cohesive Devices as Textuul Construints on Relevunce 149
However, on closer inspection of (5), it is precisely this part of the text that contributes
new propositional meaning in the coherence unit. Al1 the rest is metadiscourse material. On the
one hand, there is the NP the water, whose metadiscourse function is of a textual type, since it
encapsulates backwards. On the other hand, there is the NP containing the great scary ocean,
which is part of a metaphorical phrase used by the writer to show her attitude towards the new
propositional meaning conveyed in the post-modifier. Thus, this metaphorical phrase also
contains metadiscourse material of an interpersonal type. The process of encapsulation at (6)
seems to have operated at least on the new propositional meaning generated by interpreting (5).
In the method section, 1have already discussed the point-to-point cohesive effect of ellipsis such
as the two cases in (23) in relation to (21) in the sense that the elliptical material recovered only
refers to one part of the wording used to express the new propositional content of the related
fragment of text, thus establishing relevance of wording.
(21) -the chief examiner always loomed over one's shoulder, checking, commenting,
re-marking if necessary. (22) At least, 1think that's what he did. (23) Even if he didn't,
the fear that he would was a great deterrent to misdemeanour.
This seems to point to the working conclusion that this type of ellipsis cannot on its own
establish true encapsulation. If encapsulation is perceived, it should be attributed instead to the
textual role of the reference personal pronoun he. From our relevance perspective, it could now
be understood that when the reader approaches the interpretation of he, slhe does not only relate
it to the phrase the chiefexaminer mentioned in (21), but to the assumption generated by the
interpretation of the whole of (2 1-22), along the following lines:
(23) Even if he [the chief examiner whom at least 1 think always loomed over one's
shoulder, checking, commenting, re-marking if necessary] didn't *[loom over one's
shoulder, checking, commenting, re-marking if necessary], the fear that he [the chief
examiner whom at least 1 think always loomed over one's shoulder, checking,
commenting, re-marking if necessary] would *[loom over one's shoulder, checking,
commenting, re-marking if necessary] was a great deterrent to misdemeanour.
1presume that the same may occur with substitution, though 1cannot find clear evidence
in this text. The only possible case is do in (60). However, this is an ambiguous case, since it
might be considered as an emphatic auxiliary do followed by ellipsis o r a true substitute. In any
case it would be canying out point-to-point cohesion. Another point-to-point cohesive signal in
this clause would be those, which -together with do- is by the way the only case of point-to-
point cohesion identified by the subjects as really textual. In this coherence unit, though, the
most powerful coherence mechanism perceived in relation to previous discourse was the explicit
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 ( l ) , 2003, pp. 1 11-165
encapsulation established by the adversative logical signal but, as well as the fact that this clause
was also contributing to satiseing the prospection created in (58).
A different case, which should not be confused with substitution, is what happens with
the general verb do, as in (22). This general verb can be considered as a general lexical word,
which in the combination he did that manages to encapsulate the semantic meaning of the whole
of (21) by means of deictic acts that establish relevance of content. In any case, again the most
powerful coherence mechanism perceived in (22) is the explicit encapsulation effected by the
adversative corrective signal a t least. And if we look at the data in detail, there is no single
coherence unit whose relevance is perceived through point-to-point cohesion signals only. If
these do occur, there is always a more powerful device at work to establish relevance, as in (60).
Another case of combination of devices that is not considered by Halliday and Hasan (1 976), for
instance, is what we find in (24) in relation to (2 1-23):
(24) But how much simpler and more thorough * (than the chief'examiner looming over
one S shoulder ...) is the retuming of marked scripts to the original writers.
The authors analyze the occurrence of the comparative forms -er and more as cases of
indirect anaphoric reference. They state that "particular comparison, like general comparison,
is also referential; there must be a standard of reference by which one thing is said to be superior,
equal or inferior in quality and quantity" (Halliday & Hasan, 1976: 81). And in fact their
referent, though indirect, is easy to identi6 in the previous text, (2 1-23). What 1 would like to
claim is that this type of comparison also involves the phenomenon of standard ellipsis (cf. Quirk
& Greenbaum, 1985: 888), and that this type of ellipsis is really textual since it helps to establish
relevance of content encapsulating the whole of the meaning created by a previous fragment of
discourse. It is not merely establishing relevance of wording.
VI.4. Prospection
It is interesting to note that the textual role of none of the coherence units in the text structure
can be accounted for exclusively on the basis of prospection. In al1 cases prospection co-occurs
either with encapsulation, as in (2), or encapsulation and inferred encapsulation as in (28), or
inferred encapsulation only, as in (7) and (47). This seems logical as the places where
prospection might be expected as the exclusive coherence mechanism would be in the first
coherence unit of a text, or in the first unit of a totally independent segment of the text, as may
happen when the subject changes completely (cf. Sinclair, 1993: 14). However, prospection
occurred in this text at points where it was also possible to perceive a connection between the
prospecting coherence unit and previous discourse by virtue of some encapsulating device,
whether explicit or not.
The Role of Cohesive Devicrs os Texi~talConslroinls on Reltvonce 15 1
The only remarkable feature worth mentioning is that there have been cases where the
prospecting word to be elucidated in upcoming discourse is not expressed in a plural form. It is
simply presented as new to the context in an indefinite nounphrase, such as an evaluation in unit
(7). According to the group interpretation, on reading coherence unit (7), the writer seemed to
be committed to specifying the contents of such an evaluation in the following text. In fact,
coherence units (9-1 1) were perceived as satisfjing the prospection created in (7).
It is worth pointing out that from a text-as-product view it might be said that coherence
units (26), (34), (41), and (47) also serve to satisfj the prospection created in (7). However, when
the subjects had to analyze the relevance of these new units they had forgotten that a prospection
had been created in (7). In any case, they were able to connect them with other units in the text
by means of encapsulation. For instance, the students were able to perceive encapsulation over
(6) at coherence units (26) and (47). It should be recalled that unit (6) introduces the major topic
of the text, the pilot scheme. Thus, although the subjects could not state that these coherence
units really fulfilled the prospection generated at (7), they did perceive the fact that both
coherence units were conveying a form of evaluation of the pilot scheme.
V1.4.1. The role ofquestions in the text

Questions have been considered as the most obvious ways of creating prospections (cf. Sinclair,
1993: 12). And this is true in this text too. For example, the one question that clearly does this
is in coherence unit (2), andyou know what? From our relevance-based framework, it might be
said that the question may have been introduced to establish the relevance ofthe following piece
of discourse (3-4) by means of the prospection that it creates. This could also be interpreted as
an attempt to engage the reader in the reading process, but it was clearly understood by the
subjects that it would be the writer who would fulfil the prospection created.
However, not al1 questions in the text are used in this way. For instance, coherence unit
(8), how was itfbr you?. was expressed as a question too. But it was obvious that in this type of
written communication the reader could not respond immediately to the writer's message. Nor
could the writer answer for the reader. Thus, it cannot be said that a prospection is created at this
point. What is interesting to note is that this question is placed after a unit that does create a
prospection, (7), but (8) does not serve to fulfil it. In cases like this we are faced with a
phenomenon that may be temed interpolation in the sense that the interpolated sentence occurs
after a prospection but is not at least part of the fulfilment of the prospection. Therefore, the
prospection created in the previous unit remains (cf. insertion sequences in Schegloff, 1972). In
fact, readers need to wait until coherence unit (9) to see at least part of the prospection fulfilled.
The interpolated question at (8) is a novelist's cliché said by one of the partners after an act of
sexual intercourse, and is here used to make a silly journalistic joke. It creates the assumption
that the reader of this text had participated in the same pilot scheme as the writer is describing.
This may be perceived as an attempt by the writer to engage the readers more actively in the
communication process by asking them to evaluate their own experience of the pilot scheme. It
152 Ana I ,ifot.eno
is interesting to note how the writer adapts a cliché from the same domain "did the earth move
for you?" at coherence unit (10) to restate the general evaluation of the pilot scheme made at
coherence unit (9). which clinches the intention of the writer at coherence unit (8).
On the other hand, both coherence units (16) and (5 l), though expressed in the form of
a question, were understood as rhetorical questions. i.e. forceful statements which had the form
of a question but which did not expect an answer. They were functioning as negative statements
rather than questions. Thus, for instance. (16) might be reformulated as I would not say such a
thing. Likewise, in coherence unit (51) the question should they be returned at all? was
understood as a negative statement meaning they should not be returned at all.
VI.5. Fulfilment of prospection + encapsulation

Twelve coherence units in the text serve to fulfil a prospection (20.33%). In almost half of these
cases, though, encapsulation seems to coexist with the fulfilment of a prospection, such as in
coherence units (4), (1 l), (30), (48) and (60). If this were really the case, we would have found
evidence to contradict Sinclair's claim (1993: 12) that "a sentence cannot simultaneously fulfil
a prospection and encapsulate the utterance that makes the prospection. The former requires
maintenance of the discourse function of the previous utterance, and the latter requires the
cancellation of that discourse function". However, on closer inspection, we shall realize that the
reason for this co-existence is that in the present analytical framework each of the coherence
units in a rhetorical routine fulfilling a prospection created in previous discourse has been
classified as such, i.e. as fulfilling a prospection. Sinclair, by contrast, only seems to classifj the
first sentence in the rhetorical routine satis@ing the prospection as prospected.
In any case, it is worth noting that the encapsulated coherence unit in these cases was not
the coherence unit that had created the prospection. For instance, there is additive logical
encapsulation between (3) and (4), and the two coherence units together can be said to fulfil the
prospection created in (2). So (4) was classified both as encapsulating over (3) and fulfilling the
prospection created in (2). The same happens with the following: (1 1) in relation to (6) and (7);
(30) in relation to (29) and (28); and (60) in relation to (59) and (58). But in neither case does
the scope of the prospection coincide with the scope of the encapsulation. From this perspective,
one possible way to adapt Sinclair's words so as to improve the explanatory power of this
analytical framework might be to state the following: the coherence unit (or the first coherence
unit in a rhetorical routine) fulfilling a prospection cannot simultaneously fulfil the prospection
and encapsulate the utterance that makes the prospection.
Even so. there is one exceptional case that does not seem to comply with this norm. For
example, coherence unit (47) makes a prospection over a group of sentences (48-50), which
enumerate logistical problems advanced by means of the lexical superordinate problems (an
enumerable). If we now take (48), this sentence clearly encapsulates over (47), by means of the
The Role of Cohesive Devices us Textuul Construints on Relwat7ce 153
conjunct for instance, which might be paraphrased as "an example of these logistical problems
that there will be is", whose scope is the whole of the previous sentence.
This exactly corresponds to what Tadros identified as a typical phenomenon in
prediction: "following a prediction of enumeration sequencing signals Virstly,,finully, one, next,
further, etc.) are one of the means whereby we can recognize the different heads" of the
predicted members (Tadros, 1985: 20). These cases might be considered as exceptions in the
sense that the enumerating signals do not seem to cancel the prospecting function of the
coherence unit creating the prospection. They only serve to organize the textual material used
to fulfil it. In any case, it is also important to notice that the scope of the encapsulation does not
coincide with the scope of the prospection.
Also worth noting are cases where a discourse fragment fulfils a prospection created in
a previous sentence. but encapsulates discourse generated by a different piece of text to that
which created the prospection. For instance, coherence units (9-1 1) serve to satisfy the
prospection created in (7). However, units (9) and (1 1 ) also seem to encapsulate previous
discourse meaning -that generated by (6), at least- in both cases. So, again, although the
f~ilfilmentof a prospection and encapsulation may co-exist in the same current coherence unit,
the sentences fulfilling the prospection do not encapsulate the utterance making the prospection.
Having analyzed al1 types of textual cohesive ties in the text, and discussed cases of difficult
classification. table 4 summarizes the major mechanisms identified in the text as truly textual,
bearing in inind that some of them occur in further combinations, as table 3 shows above.
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES, vol. 3 ( 1 ) . 2003. pp. 1 1 1-165
154 A n a I. X101.eno
TubIe 4: Textual mechanisms of coherence identified in the text

Coherence
Subtype
mechanism
Encapsulation
Relevance of Deictic Reference Personal
content acts Demonstrative
Comparative
Phrasal: Demonstrative + Repetition

Reference + Demonstrative + Repetition with word class change
Lexical + Synonym + Repetition
Demonstrative + Superordinate
Synonym + Synonym with word class change +
Synonym
Comparative + General word
Personal + General word + Demonstrative
Superordinate
Clausal Same meaning (Demonstrative + Personal +

General verb)
Similar meaning
Inferred meaning
Opposite meaning
o Situation
o lnstance
Discourse Phrasal Demonstrative + Superordinate

acts Reference +
Lexical
Relevance of Logical Additive Positive
relational acts Avowal
fiinction Exemplificatory
--
Causal (Cancellation of) lnferred consequence
Conditional
Reversed polarity conditional
Adversative Proper
Corrective
Contrastive
The Role of Cohesive Devices us Tex/~rulCons/ruin/soti Relevunce 155
lnferred encapsulation
Relevance of Deictic Elliptical Post-modifier
con ten t acts Subject + Operator
Comparative clause
Clause (other)
Discourse Elliptical Post-modifier

ac 1s
Relevance of Logical Additive
relational acls Expository:
function Metaphorical paraphrase
Paraphrase
Causal Claim-support
Claim-reason
lnferred consequence
Adversative Proper
Dismissive
Contrastive
Encapsulation + inferred encapsulation

Relevance of 1 Deictic Reference + Comparative word + Comparative clause
-- --. -- -b t s
content Elliptical --
Encapsulation + inferred point-to-point cohesion
Relevante of Deictic act / Clausal: Personal + Predicator ellipsis
content + + Wording 1 Reference
relevance of ( act 1
+ Elliptical
wording 1 1
Encapsulation + prospection
Relevance of 1 Logical Causal Action-reason
relational acts
function
Prospection
Relevance of 1 Deictic Reference Question word
1 content acts
Punctuation Question mark
Colon
Superordinate
Demonstrative + Superordinate 1
1 1 Reference + 1
Lexical
Comparalive
Fulfilment of prospection
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 ( 1 ), 2003, pp. 1 1 1- 165
VII. CONCLUSIONS
The present paper has determined which features of a given text have an important role in
helping a discourse community of undergraduate student subjects to perceive the relevance and
coherence of the text in the process of reading. It has also shown how these expressions serve
both the writer and the readers as textual constraints to optimize relevance in accordance with
the Principle of Relevance. The results clearly show that there are many cases of point-to-point
cohesion that cannot be regarded as textual in nature in the sense that they were not essential to
account for the relevance of each successive text ofthe moment. By contrast, there are a number
of cohesive resources that deal only with discourse meaning derived from entire sentences, larger
fragments of text or. occasionally, certain simple clauses linked paratactically, and they do much
more than effect a tenuous connection between isolated constituents of sentences. This validates
the reformulation of Sinclair's (1993: 19) hypothesis about text structure.
What seems clear is that in the discourse perceived from this text by the group of subjects
at least one coherence mechanism was identified to relate every current coherence unit to its co-
text. There was in most current units at least one encapsulating mechanism, whether explicit,
inferred, or both. As well as this, there were very few cases of point-to-point cohesion. However,
these did not seem to account for relevance by themselves. There was always a more powerful
mechanism to account for coherence at that point. There were also a number of prospections.
The only current coherence units where encapsulation did not occur were the following: the first
sentence in the text and those fulfilling a prospection. In the latter cases, however, this norm only
seemed to apply consistently to those coherence units that initiate the rhetorical routine satisfiing
the prospection because there were some encapsulations between the coherence units within the
same rhetorical routine.
The metadiscourse items identified or inferred, which are part ofthe interactive apparatus
of the language, may be allocated a textual role in that they serve to give independence to the
sentence, and occasionally the clause. They helped to progressively determine the status of
previous or upcoming text in relation to the current coherent unit (cf. Sinclair, 1993: 7). That is
why it is possible to confirm that these metadiscourse elements serve to give independence to
a coherence unit and also help to perceive it as relevant.
The results from this study support Sinclair's suggestion that considering the pure
orthographic sentence, i.e. the clause complex enclosed by a full stop, as the minimal unit for
text structure may have to be revised slightly. The present analysis identified clauses in
paratactic relationships within several clause complexes that proved to be autonomous from a
coherence point of view. They were separated from a neighboring clause by a colon, a dash, or
a comma or dash followed by some cohesive device. However, the analysis of the connection
between these clauses and previous or upcoming discourse identified logical acts establishing
relevance of function andlor deictic acts, or discourse acts, establishing relevance of content,
usually in connection with the meaning conveyed by the other clause in the same clause
The Role o f Cohesive Devices as Texfual Consfroinfson Rt.ltvanct. 157
complex. In this sense, the full stop should not be considered as the only adequate indicator of
a coherence unit. There may be other indicators of a coherence unit, such as the colon, the dash,
or the comma or dash followed by some cohesive device that might also be adequate. The
question to be elucidated is in which circumstances this may happen.
If we accept this redefinition of minimal coherence unit, which does not correspond
strictly with the orthographic sentence, it might be more adequate to relate the concept of text
of the moment -which is very useful from the point of view of discourse interpretation-
directly to the concept ofminimal coherence unit, instead of relating it to the abstract theoretical
construct of sentence. The problem again is to find out how the minimal coherence unit is
enclosed in actual written discourse. Further studies should then focus on this definition on the
basis of more evidence. However, this is not bad news. As Hyde puts it: "Discourse analysis is
(for the moment at least) not so much interested in the definition of a sentence in structural,
system terms as investigating where and why writers place full-stops (in written texts)" (Hyde,
1990: 188). Why not colons, semicolons, dashes, and commas followed by some cohesive device
too? "This point of view focuses more on chunking than on structure. that is to say, on how much
information is loaded on to a given format" (Hyde, 1990: 188).
As has been noted in the discussion, in order to establish relevance, meaning was not
always derived from the immediately preceding coherence unit only but also from other parts
of the text. In view of the data, 1 would suggest reformulating Sinclair's hypothesis by saying
that the encapsulation process does not operate necessarily only over the previous sentence, or
rather the meaning created by it, but over the most relevant information stored in the reader's
mind. In written discourse, as we have seen, the most relevant information is usually the
information derived from the previous coherence unit which is stored in the reader's short-term
memory. But the text has shown examples where the relevant fragment of text from which
meaning was retrieved was a larger immediately preceding fragment, although the immediately
preceding coherence unit was at least part of the referent, such as (26-33) in connection with
(34). On the other hand, the study has also shown that, in order for the subjects to perceive
relevance, meaning was also derived from other parts of the text which had been stored
somewhere in the reader's short-term memory, such as (26) and (5 1) together with (53-54) in
connection with (55).
The only problems that arise in terms of the encapsulation hypothesis are at coherence
units (41) and (47), which are both the beginning of a paragraph. In both cases, it was quite
obvious to the subjects that there was no connection with the immediately preceding sentence
or even paragraph. At (4 l), a connection could be established in relation to coherence units (26-
33) in the sense that it presented an unexpected result in relation to the meaning developed in a
previous but not immediately preceding paragraph. At coherence unit (47), the connection was
established with sentence (6). Therefore, in this case the encapsulation neither scoped over at
least the immediately preceding coherence unit or fragment.
158 A n a f. ~bforeno
It could be argued that the problems here do not really have to do with the analytical
scheme but with the subjects' capacity to perceive the text structure. The point is that, for
whatever reason, the subjects had forgotten that a clear prospection had been established at
coherence unit (7), where it was announced that an evaluation of the pilot scheme was going to
be made. Had they maintained that prospection open throughout the text, they would have
perceived some of the upcoming paragraphs as fulfilling that prospection. Looking at the text
from hindsight, it could be interpreted that the paragraphs starting at coherence units (9). (26),
(34), (41), and (47) fulfil such a prospection in the sense that they present evaluative material
about the pilot scheme introduced at coherence unit (6). And that may account for the
discontinuity of the encapsulation process at some of these places, some of which were found
problematic. One reason why the subjects might have forgotten the prospection may have had
to do with the fact that the analysis was carried out in various sessions for unavoidable practica1
reasons. Thus when the subjects approached the text in subsequent sessions their short-term
memories may have been weakened. A clear implication for further studies of this kind is that
the text used should be shorter so that it may be dealt with in just one session.
The model proposed by Sinclair (1993) was based on one single text, just as this is, and
it may need some refinement as more discourse units and genres are analyzed. However, 1
believe that this model is nowadays the most explanatory in accounting for those text elements
that contribute to our perception as readers of discourse coherence and structure. It may also be
applicable to other types of text with little adaptation. Thus table 4, which does not intend to be
exhaustive, might serve as a good basis on which to build the model further. More importantly,
this model is highly consistent with a cognitive view of discourse interpretation. This, in my
opinion, gives it even more support, because a cognitive view of the role of cohesive items in
the perception of relevante, and therefore coherence, seems much more convincing than
previous accounts.
Lastly, the reader of this article may not have perceived the coherence of the text that we
have analyzed in the same way as has been reported here. But that does not invalidate my
argument because that is likely to have happened. As has been suggested, there may be multiple
discourses from a single text. 1would invite the readers of this paper to try and do the same test
to find out which text features help them to make sense of the text at every stage and to check
whether their interpretation is similar or differs from the one reported here. What 1 can
guarantee, after my experience of using the outlined model of text structure and coherence with
advanced undergraduate students of English Philologv and doctoral students for about five years
now, is that they find it highly explanatory and convincing. What is more important, students
consider it as very clarifying in their understanding of coherence and how interpretation of
written discourse works.
The Role of Cohesive Devices us Textual Construints on Relevante 159
Acknowledgements
I would like to thank John Sinclair and John Hyde for theircareful readings and insightful comments on previous
versions of this paper. Any infelicities that remain are, of course, my own. 1 would also like to express my
gratitude to the students at the Department ofFilologíu A4odernu (University of León, Spain) that participated
either in the pilot or the final study for their constant enthusiasm and invaluable comments during the
interpretation process.
REFERENCES
Blakeiiiore, D. (1987). Semantic C'onstraints on Relevance. Oxford: Blackwell.
Blakeiiiore, D. (1992). Understanding Utterances. An lntrodtlction to Pragmatics. Oxford: Blackwell.
Brazil, D. (1995). A Grammar ofspeech. Oxford: Oxford University Press.
Cook, G. (1 989). Discourse. Oxford: Oxford Uiiiversity Press.
Croinbie, W. (1985). Process and Relation in Discoil~seand Language Learning. Oxford: Oxford
Uiiiversity Press.
Dowiiiiig, A. & Locke. P. (1992). A University Cotrne in English Grammar. Hertfordsliire: Preiitice Hall
liiteriiational (UK) Ltd.
Grice, H.P. (1975). Logic aiid coiiversation. 111P. Cole & J. Morgan (Eds.), Syntax and Semantics. vol.
3: Speech Acts (pp. 41 -58). New York: Academic.
Hal liday, M. A. K. ( 1 973). Explorations in the Ftlnctions ofLangtiage. Loiidoii: Edward Ariiold.
HallidayoM.A.K. & Hasaii, R. (1976). Cohesion in English. London: Loiignian Group Limited.
Hyde, J. (1 990). Aspects ofDiscoirrse Analysis: The Explicit Signalling of 1ntc.isentential Relations in
English. Uiipublislied Doctoral Dissertatioii. Uiiiversity of Salainaiica, Spain.
Hyde, J. (2002). Soine basic issues in tlie study of laiiguage today: An iiitroductioii to the aiialysis of
iiiterseiiteiitial relatioiis in Eiiglisli aiid Spanisli iiewspaper editorials. 111P. Aloiiso, M.J. Sancliez,
J. Hyde & C. Moran (Eds.) Aspects of Discotrrse Analysis (pp. 2 1-4 1). Salamanca: Ediciones
Uiiiversidad de Salainaiica.
160 Ana I. ~Woreno
Mann, W.C. & Thonipson, S.A. (1 986). Relational propositions in discourse. Discotirse Processes, 9. 57-
90.
Moreiio, A. 1. (1 996). Esttidio contr-astivo inglis-espaiiol de la expresión de las relaciones de coherencia

causal interoracional: El artíctilo acadimico sobre Economía y Empresa. Leóii: Secretariado de
Publicacioiies de la Universidad de Leóii (CD electronic forniat: Tesis doctorales IYYS, vol 26).
Moreiio, A.1. ( 1997). Geiire coiistraiiits across laiiguages: Causal inetatext in Spanisli and Eiiglisli. ESP
Journal, 16:3, 161-1 79.
Moreno, A.I. (1998a). A siinple inodel of text structure: Tlie aiialysis of aii ESP text. I i i S. Barrueco, E.
Henáiidez & L. Sierra (Eds.), Lengtias para Fines Especi~cos.(VI) ( p p 275-283). Alcalá de
Heiiares: Universidad de Alcalá.
Moreno, A.1. (1998b). Tlie explicit sigiialliiig of prernise-coiiclusion sequences in research articles: A
contrastive frarnework. Texr, 18:4, 545-585.
Moreno. A.1. (In press). Matching theoretical descriptions of discourse and practica1 applications to
teacliiiig: Tlie case of causal rnetatext. English for SpeciJic Purposes, 22:3,265-295.
Moriarty, H. (1999). Exain scripts pilot gets top tnarks for effort. Gtra~.dian Unlimited.
Iittp:l/educatio~i.guardiaii.co.uk/old/s/O, 10044,10655 1 ,OO.htnil
Nuttall, C. (1 982). Teaching Reading Skills in a Foreign Langtiage. Oxford: Heiiieinaiiii.
Pérez Álvarez, 1. (2001). Aplicación einpirica del inodelo de estructura textual de Sinclair (1993). Iii A.
Moreno & V. Colwell (Eds), Recent Perspecrives on Discou~*se.Leóii: AESLAISecretariado de
Publicacioiies y Medios Audiovisuales. Uiiiversidad de Leóii (CD electronic foriiiat).
Quirk, R., Greenbaurn, S., Leech, G. & Svartvik, J. (1985). A Comprehensive G i ~ m m a of

r [he English
Language. New York: Longinaii Group Ltd.
Sanders, T.J.M., Spoorem, W.P.M. & Noordrnan, L.G.M. (1993). Coherence relations in a cognitive
theory of discourse represeiitation. Cognitive Linguisrics, 4:2, 93-1 33.
Siiiclair, J.M. (1992). Trust the text. I i i L. Ravelli & M. Davies (Eds.), Advances in Systemic Linguistics:
Recenr Theory and Practice (pp. 5-1 9). Loiidoii: Pinter.
Siiiclair, J.M. (1993). Written discourse structure. In J.M. Siticlair, M. Hoey & G. Fox (Eds.), Techniqties
of description: Spoken and Written Discourse. a Festschrijl for- Malcolm ('oulthard (pp. 6-31).
Loiidoii: Routledge.
Sperber, D. & Wilsoii, D. (1986). Relevante: Communicarion and Cognition. Oxfored: Blackwell
O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES, vol. 3 (l), 2003, pp. 1 1 1-165
The Role ofcohesive Devices as Textuul Construints on Kelwance I hl
Tadros. A. (1985). PI-ediction in Text (Discotrrse Analysis Monographs, 10). Birmingham: English
Laiiguage Researcli. Universiiy of Birmiiighain.
Tadros: A. (1994). Predictive categories in expository text. In M. Coulthard (Ed.), Advances in Written
Text Analysis (pp. 69-82). Loiidoii: Rouledge.
Vaiide Kopple, W.J. (1985). Soine exploratory discourse oii inetadiscourse. College Composition and
C'ommunication, 36,82-93.
Williains, J.M. (1981). Style: Ten Lessons in Clai-ity and GI-ace.Glenview, IL: Scott. Foresman and
Coinpaiiy.
Wiiiter. E.O. (1977). A clause relatioiial approacli to English texts: A study of some predictive lexical
items in writteii discourse. Instructional Science, 6:1, 1-93.
Wiiiter, E.O. (1986). Clause relatioiis as iiiforinatioii structure: two basic text structures in Eiiglisli. 111R.
M. Coultliard (Ed.), Talking abozrt Text (Discozrrse Analysis Monographs. 13) (pp. 88-108).
Biriniiigliain: Eiiglisli Language Researcli. University of Birmingham.
8 Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES, vol. 3 ( 1 ), 2003, pp. 1 1 1 - 165
162 Ana l. Moreno
APPENDIX
Segmented text
Exam scripts pilot gets top marks for effort
The verdict on returning exarnination papers to students? Fairly good, roorn for
irnprovernent
Hilary Moriarty
Tuesday November 23, 1999
The Guardian
( 1) Nineteen ninety-nine was the year we dipped a toe in the water:
(2) and you know what? <
( 3 ) [The sharks didn't bite,
(4) and the water wasn't freezing.]
(5) The water was the great scary ocean of returning examination papers to candidates.
(6) This year saw the pilot scheme, with three different models for GCSE and at A leve],
for the copying and return of al1 scripts in 1O svllabuses, allowing centres to decide
how to release the copied scripts to candidates.
(7) The Qualifications and Curriculum Authority has carried out an interim evaluation
* (of the pilot scheme). <
(8) "How was it for you?"
(9) [The great news * (about the pilot scheme) is that there seems to be general
a u ~ r o v afor
l the vrinciple of returninq the scripts.
(10) * (In other words) The earth rnay not have moved, but the world didn't come to
a standstill either.
(1 1) It was E . ]
(12) * (In fact) Not surprisingly, most of the peoule involved * (in the pilot scheme)
felt that returning the scripts made the examination svstem more transparent and
exaininers more accountable.
The Role o f Cohesive Devicev os Textuol Consboints un Relevonce 163
(1 3) * (because) Sometimes you don't need to te11 peovle to work better, you just te11
them there's an audience for what they produce.
(14) * (In other words) Knowing that whatever was done to the Dauers would be seen
in the outside world must have been salutary.
(1 5) This is not to say that examiners were sloppy before.
(16) Would 1say such a thing? (= 1 would not say such a thing)
(1 7) * (because) 1examined for years:

(1 8) * (examining was) the most gruelling in the world, requiring painstaking
effort and concentration to sustain standards justly for 300 s c r i ~ t in
s three weeks.
(19) * (In other words) Conscientious marking is a m.

(20) * (And = but) And examiners never did &in an irresponsible vacuum -
(21) * (because) the chief examiner always loomed over one's shoulder. checkinq,
commenting, re-marking if necessary.
(22) At least, 1think that's what he did.
(23) * (Anyway) Even if he didn't, the fear that he would was a great deterrent to
misdemeanour.
(24) But how much simpler and more thorough * (than the chief examiner looming
s the original writers.
over one's shoulder ...)is the retuming of marked s c r i ~ t to
(25) * (Returning the marked scripts... is) Real accountability.

(26) T h e irony * (of the pilot scheme) is, of course, that * (in spite of) having been
offered their scripts, most of the candidates didn't want them.
(27) * (As a matter of fact) Staff in the centres reported the percentage of students
"very interested" in viewing the s c r i ~ tas
s about 12%, with a further 27% only "fairly
interested".
(28) The reasons * (why most of the students did not want to view the scripts) are
obvious: <
(29) [if you did well, you really don't about the uapers -
(30) and that goes for doing well unexpectedly, as well as having the satisfaction of
achieving just what you expected.]
O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES. vol. 3 ( l ), 2003, pp. 1 1 1 - 165
164 Ana f. ,bfoteno
(3 1) * (By contrast) Interest in the papers is generated by doing badlv,

(32) * (and = but; then = that) and then only if it surprises you.
(33) * (because) If you partied al1 year, or had a personal crisis, then you will have
done badlv but you won't need to see the uauers to see why.
(34) The interim revort indicates also that DuDils needed teachers to decode what they
m-
(35) * (this is) small wonder, if the rumours are right and examiners were virtually
forbidden to write on the scriuts for fear of litigation from insulted students.
(36) * (because) Without some sort of written exvlanatory commentary, candidates

might well find the s c r i ~ t"more
s meaningful when interpreted by their teacher".
(37) Actually, if the pilot scheme is judged successful and more s c r i ~ t are
s returned
in the future, this is an area where practice must be improved.
(38) * (because) Particularly in arts subjects, where markingis notoriously subjective,

the examiner's commentary is vital evidence.
(39) * (In fact) In my day, 1 was expected to annotate scripts to explain my marks to
-the chief examiner.
(40) Remove that requirement, and & examining urocess will only appear to be
more m,while in fact retaining an almost smug inscrutability.
(41) * (If = While) If candidates didn't care about the scripts, 71% of staff cared
a great deal:
(42) * (As a matter of fact) 82% _* agreed that access to the scriuts would help with
teachinn the svllabus in the coming year.
(43) * (Of course = this is natural) Well of course.

(44) * (because) Knowing exactlv where the last candidates got it wrong is the best
learning tool a teacher can have to improve performance next vear.
(45) * (However) Better than knowing what they got is knowing why they g- it.
(46) * (So) If any g o v e m e n t wants to conjure up massive whole school
im~rovement,this is the magic wand.
(47) There will be logistical problems * (with the process of returning the scripts
to candidates): <
(48) [returningall scriuts will mean 13.5m papers whizzing through the postal system,
for instance.
(49) Photocowing scriuts sounds horrendous even to a convinced "pro-retumer" like

me.
(50) Proper scrutiny of the pauers in school will take time, possibly precious holiday
time.]
(5 1) And if the big learners here are teachers, not pupils, should they be returned
at all? * (= with al1 these problems, it looks as if they should not be retumed at all)
(52) * (However) The answer (to this question) is yes (they should be returned).
(53) * (because) 1believe now, as 1 believed last year when 1 wrote one of the first
articles calling for this move towards long-overdue transuarencv and accountabilitv,
and as the authorities hold in New Zealand, that it is simply the right thing to do.
(54) * (And) The right thing overrides logistical problems.

(55) Pupil neglect of the papers is beside the point. * (= is not relevant to the
question)
(56) * (because) A few _* will be very interested indeed,

(57) and that's enough.
(58) * (It is) A bit like voting, really: <

(59) [lots of peoule don't m about that either,
(60) but for those * (people) who do * (care) , it's one of the markers of a civilised
world.]

Dialnet TheRoleOfCohesiveDevicesAsTextualConstraintsOnRele 713368 PDF

Uploaded by

Copyright:

Available Formats

Dialnet TheRoleOfCohesiveDevicesAsTextualConstraintsOnRele 713368 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dialnet TheRoleOfCohesiveDevicesAsTextualConstraintsOnRele 713368 PDF

Uploaded by

Copyright:

Available Formats

International Journal

The Role of Cohesive Devices as Textual Constraints

KEYWORDS: discourse, text, reading process, coherence, relevance, cohesive devices,

11. THEORETICAL FRAMEWORK

Given the role of inference in establishing the contextual effects of a proposition,

111. DESIGN OF THE EMPIRICAL RESEARCH

111.1. Research strategy

111.2. The subjects

111.3. The data

O Servicio de Publicaciones. Universidad de Murcia. Al1 rights reserved. IJES,vol.3(1),2003,pp. 111-165

of the original text.

111.4. The procedure: a discourse-as-process view of the text

the first unit of coherence for focussed attention.

Table 1: Sample Test Items (1 to 2)

Nineteen ninety-nine was the year we dipped a toe in the water:

and you know what?

IV. METHOD OF ANALYSIS

IV.1. Encapsulating, or retrospective, mechanisms (E)

Ir 1.I . Relevance of content

IV. l . 1.a. Encapsulating deictic act

In my view, it is possible to appreciate a mechanism of encapsulation as defined by

IV. l . 1.b. Encapsulating discourse act

IV. 1.2. Relevante of relationul jknction (Encapsulating logical act)

IV.2. Prospective mechanisms (P)

IV. 2.1. Relevunce of'content

IV.2.1 .a. Prospecting deictic act

IV.2.1.b. Prospecting discourse act

After reading (58), it seems as if the writer is committed to performing an act of

IV. 2.2. Relevance of relational function (Prospecting logical act)

IV.3. Units fulfilling or satisfying a prospection (S)

IV.4. Inferred encapsulation, or qualified assignments (1)

ZV.4.1. Relevance ofcontent (Znferred encapsulating deictic act)

IV. 4.2. Relevance ofwording (Inferredpoint-to-point wording act)

IV.5. Type of cohesive tie

IV.6. A text-as-product view of cohesive devices

P: Deictic Reference: Question word

2) and E: Logical Additive: Positive

Additive: Expository: Paraphrase

1 Clausal: Personal referente +

matter of 1: Logical Causal: Claim-support

2) and E: Logical Additive: Positive 29

33 1 2) and * (and = but)

O Servicio de Publicaciones. Universidad de Murcia. All rights reserved. IJES,vol.3(1),2003,pp. 111-165

best learning tool a

1: Logical Causal: lnferred consequence 45

Tuble 3: Major coherence mechanisms affecting each coherence unit

tulfilniciit o f prospcctioii + Iiifcrrcd E. 2 34 9. 10

VI.l. Inferred encapsulation

A smaller number of inferred encapsulations have been perceived in combination with

VI. 2.1. Encupsuluting clausul deictic uct

VI.2.1.a. Similar meaning (similar meaning clause)

In (41) the propositional content expressed by the seemingly conditional clause

VI.2.1 .b. Inferred meaning (Inferred meaning clause)

VI.2.1 .c. Opposite meaning (opposite meaning clause)

VI.3. Textual versus Point-to-point cohesion

V1.4.1. The role ofquestions in the text

VI.5. Fulfilment of prospection + encapsulation

TubIe 4: Textual mechanisms of coherence identified in the text

Phrasal: Demonstrative + Repetition