Writing Without Conviction ?

Hedging in Science
Research Articles
KEN HYLAND
English Department, City University o} Hong Kong

Hedging is a well-documented feature of spoken discourse as a result of its role

in qualifying categorical commitment and facilitating discussion Its use in
academic writing has received less attention, however, and we know little about
the functions it serves in different research fields and particular genres Hedging
is a significant communicative resource for academics since it both confirms
the individual's professional persona and represents a critical element in the
rhetorical means of gaining acceptance of claims Hedges allow writers to
anticipate possible opposition to claims by expressing statements with
precision, caution, and diplomatic deference to the views of colleagues Based
on a contextual analysis of 26 articles in molecular biology, this paper argues
that hedging in scientific research writing cannot be fully understood in
isolation from social and institutional contexts and suggests a pragmatic
framework which reflects this interpretive environment

I INTRODUCTION
Hedging is the expression of tentativeness and possibility and it is central to
academic writing where the need to present unproven propositions with caution
and precision is essential Hedging has received a great deal of attention in
conversation analysis where devices such as / think, sort of maybe, and possibly
•are frequently used to create conviviality, facilitate discussion, show politeness,
and oil the phatic wheels (eg Holmes 1984,1995, Coates 1987) Hedges have
also been associated with conveying purposive vagueness (eg Powell 1985,
Dubois 1987, Channell 1994), treated as a form of metadiscourse (e g Vande
Kopple 1985, Cnsmore, Markkaned, and Steffensen 1993), and as a means of
achieving distance between a speaker and what is said (e g Prince, Frader, and
Bosk 1982, Skelton 1988) In linguistics, hedging has been the subject of a
considerable body of conceptual and empirical research and we now know a
great deal about the semantic and formal aspects of epistemic devices,
particularly the modal auxiliaries (Lyons 1977, Coates 1983, Palmer 1990)
Analyses of written academic corpora have revealed some of the character-
istics of hedging in textbooks (Myers 1992), economic forecasting (Pindi and
Bloor 1986), science digests (Fahnestock 1986), abstracts (Rounds 1982),
medical discourse (Salager-Meyer 1994), and molecular genetics articles
(Myers 1989) Stu'dies have also shown the importance to academic discourse in
Applied Linguistics, Vol 17, No 4 C Oxford University Press 1996
434 HEDGING IN RESEARCH ARTICLES

general of modal verbs (Hanania and Akhtar 1985, Butler 1990), imprecise
numenc expressions (Dubois 1987, Channell 1990) and 'commentative' items
(Adams-Smith 1984,Skelton 1988) Together, this literature has demonstrated
the clear pragmatic importance of hedging as a discoursal resource for express-
ing uncertainty, scepticism, and open-mindedness about one's propositions
However, hedging in scientific research writing represents a little-stud led area
of pragmatic competence and we still know little about how it functions or is
typically realized in specific academic domains In particular, greater attention
needs to be paid to the fact that hedging represents a writer's attitude within a
particular context There is, therefore, a need for an explanatory framework
which accounts for its pervasiveness in academic discourse by situating hedging
in its socio-pragmatic contexts
The most cogent attempt to place hedges within a comprehensive functional
structure is that of Myers (1985, 1989) who argues that hedges are part of a
wider system of politeness designed to redress the threat research claims
contain to the 'face' of other scientists While writers seek to gain recognition in
their field by making the strongest claims they can, such claims are likely to
challenge existing assumptions of the discipline and undermine colleagues'
research agendas A vanety of devices are therefore employed to mitigate
claims and minimize these impositions Myers' work is clearly suggestive and
central to any discussion of hedging, but his extension of Brown and Levtnson's
(1987) conversational model provides only a partial account of hedging in
scientific discourse The fact that the public reputation and professional
position of every scientist depends on the work and acceptance of peers means a
failure to observe appropriate norms of conduct will not merely prevent
individuals securing goals, but will incur sanctions with concrete consequences
Treating hedging as politeness thus accents the instrumental over the normative,
thereby underplaying the importance of authority and conformity in academic
discourse communities, and at the same time neglects the multi-functional
character of hedges in gaining acceptance for claims
In science, hedges play a cntical role in gaining ratification for claims from a
powerful peer group by allowing writers to present statements with appropriate
accuracy, caution, and humility Hedges help negotiate the perspective from
which conclusions can be accepted This paper provides a functional account
for the use of hedging m scientific research articles It is based on a detailed
contextual analysis of a 75,000-word corpus of 26 articles selected from recent
issues of the six leading journals in the field of cell and molecular biology (SCI
1993) A description of the major hedging expressions and their distribution in
this corpus can be found elsewhere (Hyland 1996) My purpose here is to
extend and build on previous work to propose an explanatory framework for
scientific hedging which combines sociological, linguistic, and discourse
analytic perspectives The study draws together strands from different studies to
reveal a complex overlap of motivations for hedging I suggest here that hedges
can only be understood in terms of a detailed characterization of the
institutional, professional, and linguistic contexts in which they are employed
 KENHYLAND 435

2 THE SOCIAL CONTEXT OF SCIENTIFIC STATEMENTS

The role of human judgement in data interpretation, widely accepted in the
philosophy of science (eg Kuhn 1970, Van Frasen 1980, Faust 1984, Kourany
1987), implies that a study of scientific discourse is central to understanding the
creation of knowledge Knowledge is influenced by the basic elements of the
communication process writer, audience, language, and reality This is because
transforming claims into knowledge requires reader acceptance and therefore
linguistic and rhetorical means of persuasion (eg Brufee 1986) This need for
community consensus in the ratification of knowledge, I believe, provides the
foundation for the expression of tentativeness and caution in scientific writing
statements must reflect the certainty the writer wishes to invest in them
In introducing claims, writers rely on evidential support from statements
previously confirmed by the discourse community as truths about the world,
and these can be expressed as categorical assertions Every other statement, by
which the writer asserts the content to be true as far as he or she knows, is a
hedged or non-factive statement The ratification of knowledge, of a connection
between a state of affairs used as evidence in support of a hypothesis and the
hypothesis itself, is not self-evident but depends on an audience for whatever
epistemic status it achieves Most work in normal science (Kuhn 1970), which
contributes to the cumulative extension of the discipline, is of this non-factive
character, referring to what is possibly true rather than what is certain
(1) I suggest therefore that Dl degradability must be causally linked to QB site
occupation which in turn determines PEST region accessibility to
(2) Possibly, phosphorylation of ACC synthase could contribute to the existence of its
differenl pi forms Alternatively, phosphorylation-dephosphorylation may be
involved in the inactivation/activation of the enzyme in vivo
Where there is uncertainty about the evidential status of the assumptions
between data and hypotheses, claims require varying degrees of hedging In fact,
my data indicates that few knowledge claims are presented in unmitigated form
induction and inference rather than deduction and causality characterize most
arguments in scientific discourse Thus, writers may have strong grounds for
trusting the relations between accepted premises and new results, but
uncertainty, caution, and interpersonal factors prevent the categorical assertion
of such claims In most cases, probability is explicit and negatability anticipated
Writers must acknowledge gaps and limitations in order to present claims and
conclusions in the light of accepted knowledge

3 HEDGING IN SCIENTIFIC RESEARCH ARTICLES

The publication of scientific results seeks to accomplish both institutional and
individual goals A research paper not only extends understanding of
phenomena and theories that the current paradigm deems worthy of study, but
also helps support or establish the personal reputation of the writer As I have
noted, in seeking recognition for their accomplishments writers will therefore
generally make the strongest claim for which they have epistemic authority
436 HEDGING IN RESEARCH ARTICLES

Research scientists acquire academic credibility by gaining readers'

acceptance of the most significant assertions their findings will support Secur-
ing this objective involves relating lllocutionary acts to perlocutionary effects A
wnter wants a message to be understood (an lllocutionary effect) and to be
accepted (a hoped for perlocutionary effect) But no matter how clearly,
convincingly, and appropriately reader-centred material may be expressed, the
writer's ability to influence the reader's response is severely restricted Longino
(1990) argues that there is no independent scientific framework for distinguish-
ing the observational from the theoretical and so interpretation of each change
depends on what can be contested and what taken for granted So, if'truth' does
not lie exclusively in the external world, there is always at least one plausible
interpretation of particular data Readers may therefore be persuaded to judge a
claim acceptable or may decide to reject it
This view distinguishes comprehension from interpretation (e g Perfetti and
McCutchen 1987) While lexical and syntactic forms determine text meanings,
interpretation is unconstrained and subject to knowledge effects which depend
on higher-level reasoning skills and world knowledge which are beyond the
writer's control The scientific wnter can only guide the reader to a particular
interpretation through the use of formal meanings but external factors,
particularly the reader's prior knowledge, affect interpretation (cf Bazerman
1985) Readers can always refute a claim All statements require ratification and
because readers are guarantors of the negatabihty of claims this gives them an
active and constitutive role in how wnters construct them This is why mitigation
is central to academic wnting, as hedging signals the writer's anticipation of the
opposition to a proposition
This opposition can be divided into two types, content- and reader-oriented
First, claims have to correspond with what is believed to be true in the world
Hedges here will be referred to as 'content-oriented' and concern a statement's
adequacy conditions the relationship between a proposition and a representa-
tion of reality (eg example (3)) Secondly, a proposition which could be
presented categoncally from an objective perspective may be explicitly hedged
because of reader considerations Thus, reader-onented hedges incorporate an
awareness of interpersonal factors, meeting acceptability conditions (eg
example (4))
(3) This implies that the extent to which the endophyte might effect N metabolism
under field conditions could also depend considerably upon other interacting
factors
(4) From our investigations we conclude that the data of Wydrzynski et al can be seen
in a different light when the adverse effects of CI depnvation are seen in conjunc-
tion with
In sum, hedges anticipate a need to justify claims because the wnter is
dependent on their ratification by the reader The wnter must make a
hypothesis both about the nature of reality and about the acceptability of the
hypothesis to an audience, the question of adequacy corresponding to the
 KEN HYLAND 437

objective negatabihty of a proposition and acceptability to its subjective

negatabihty
Content-onented hedges can be expressed at two further levels of delicacy
First, there are two distinct reasons for modifying a proposition so that it
corresponds with reality One is an obligation to present claims as accurately as
possible, the second concerns the need to anticipate what may be harmful to the
wnter The two are likely to overlap in actual use but forms will be referred to
respectively as accuracy-oriented (eg example (5)) and wnter-oriented
reasoning (e g example (6))
(5) These EGTA clots are transparent and possibly comprised of thinner fibres than
the
(6) The figures suggest that the determining factor for the protein response was
whether
Accuracy-oriented hedges can be further distinguished according to whether
they involve a qualification of predicate intensity {attribute hedges) or wnter
confidence (reliability hedges) The first reflect the difficulties of using a limited
language to describe the variability of natural phenomena (7), while the second
indicate writers' confidence in the certainty of their knowledge (8)
(7) Staining was generally confined to the vascular tissues
(8) The photoreceptor involved is somehow related to the photosynthetic apparatus
itself
This model is summarized in Figure 1 In summary, this framework seeks to
account for the form of sentences by showing how they anticipate possible
objections Scientists gain and keep reputations by making the highest level
claims they can, demonstrating that they deserve credit for something new But
in presenting such claims they must meet both adequacy and acceptability
conditions They therefore use hedges to reduce the risk of negation on
objective grounds, i e the match between propositional content and what reality
is believed to be like, and on subjective grounds, relating to acceptable levels of
self-assertion, deference, and willingness to debate Both are epistemically
modahzed forms of speech where what is said is related to what is meant by a
regard for reader perspective

4 A FUZZY CATEGORY MODEL

Analysis of epistemic language use soon reveals that hedging devices are
polypragmatic, they can convey a range of different meanings, often at the same
time As a result, they do not fit into a neat scheme of discrete categones which
allows one meaning to be clearly distinguished from Others Moreover, not only
is it impossible to relate particular forms exclusively to specific functions, but
the expression of simultaneous meanings introduces the problem of indeter-
minacy in specifying cases So while hedging devices express writers' attitudes to
both propositions and addressees, the choice of a particular device does not
always permit a single, unequivocal pragmatic interpretation Particular forms
438 HEDGING IN RESEARCH ARTICLES

Scientific statements

Factivc statements • Non-factive statements

Contenl-oncnied Reader-oriented

Ace u rac y-o n ented Wnter-onented

Attnbute Reliability

Figure 1 Categorization of scientific hedges

often convey more than one function and a complex overlap of usage suggests
that the precise motivation for employing a hedge may not always be clear We
cannot assume a given device will always function to hedge propositional
content, for example, nor that evidence for a particular function will always be
found in the linguistic environment The resulting ambiguity can, of course,
become an important strategic option for waters when presenting information
Consequently a theory of fuzzy-sets with graded membership, as proposed by
Zadeh (1972) and taken up by Coates (1983), appears to offer a better
approach
A fuzzy-set is a class with unsharp boundaries, that is, a class in which the transition
from membership to non-membership is gradual rather than abrupt (Zadeh 1972 4)

Zadeh observes that fuzziness plays an essential role in human cognition and
believes that most classes in the real world are fuzzy The essential feature of
fuzzy-sets is that they allow the gradual transition to membership so cases are
not restricted to binary end-points but can range between them, with some
examples denoting a greater degree of membership of one category than
another At the core, an expression will most closely approximate to the
meaning of that category while examples at the periphery will exhibit less
precise meaning Thus, core examples of wnter-onented hedges will involve
self-protection by hedging commitment, while examples toward the penphery
will show more concern with propositional accuracy
Such a model accounts for the vaganes of my corpus and recognizes the
flexibility between language and context which serves several concurrent goals
Moreover, analysis reveals the importance of four variables in determining core
cases These concern the degree of specification, venfication. agentivity, and co-
operation Each variable is related to hedging by the assumptions it allows the
reader to make about the wnter's attitude to the proposition or its context
 KENHYLAND 439

The first point concerns the extent to which the propositional content is
precisely expressed A high degree of specification of propositional elements is
associated with content-oriented strategies and particularly with attribute
hedges These determine how far the terms used accurately describe the events
or states referred to in the proposition Verification refers to the acknowledge-
ment of uncertainties about the truth of a claim It indicates the confidence that
can be invested in a statement and is associated with reliability type accuracy
hedges Agentivtty, whether the action or state described in the proposition is
explicitly associated with the writer, is cntical in distinguishing reader- and
content-oriented hedges Core examples of content hedges occur in non-writer
agent contexts, while reader-oriented cases are generally found in writer agent
environments Finally, co-operative features, such as invitations, offering
alternatives and reference to shared assumptions, indicate the extent to which
the writer seeks to involve the reader in the ratification of the claim
The model suggests the following tentative generalizations in determining
core cases

1 Where the principal role of the hedging device is to specify the extent to
which a term accurately describes the reported phenomena, it is likely to be
acting as an attribute hedge
2 Where the principal role of the hedging device is to convey the writer's
assessment of the certainty of the truth of a proposition, then it is likely to
be performing a reliability function
3 Where the device occurs in a context which conceals the writer's viewpoint
and avoids personal responsibility for propositional truth, then it is
probably acting as a wnter-onented hedge
4 Where the wnter acknowledges personal responsibility for the validity of
propositional content or invites reader involvement, then the device is
likely to be acting as a reader-oriented hedge

However, while these variables greatly assist the identification of discoursal

function in many instances, a number of them may be operative in any particular
case Indeterminacy remains a feature of hedging devices and cases assigned to
one category will often include meanings associated with another I will now
describe the model in more detail, beginning with content-orientation

5 CONTENT-ORIENTED HEDGES
Content-oriented hedges mitigate the relationship between propositional
content and a representation of reality, they hedge the correspondence between
what the writer says about the world and what the world is thought to be like
The motivations for these hedges fall into two overlapping categories, concern-
ing the writer's focus on propositional accuracy or on self-protection from the
consequences of poor judgement, although there may be an element of both
purposes on any particular occasion Accuracy-oriented and wnter-onented
hedges will be discussed in turn
440 HEDGING IN RESEARCH ARTICLES

5 1 Accuracy-oriented hedges
These involve the writer's desire to express propositions wtth greater precision
in areas often subject to revision Hedging here is an important means of
accurately stating uncertain scientific claims with appropriate caution (Rounds
3982, Skelton 1988) and aims at reducing the risk of negation on objective
grounds Almost all academic discourse is a balance of fact and evaluation as
writers try to present information as fully, objectively, and accurately as
possible Such Gncean constraints are important in research articles (RAs)
because scientific cognition is often hedged with interpretation The main
function of accuracy-oriented hedges is to imply that the proposition is based on
plausible reasoning in the absence of certain knowledge, they ask that a
proposition be understood as true as far as can be determined These hedges
represent the 'institutionalized' language of science since they help convey the
state of knowledge They enable readers to distinguish between what is actual
and what is only inferential and imply that the writer has less than full warrant
for categonal assertion Personal commitment is either not involved or is
subordinate to this function The two types of accuracy-oriented hedges,
attribute and reliability, have different motivations and realizations

5 1 1 Attnbute hedges
Science is predicated on the assumption of a world independent of language, but
the ability of words to represent a non-linguistic domain of objective facts is
heavily dependent on a shared, conventionalized cognitive schema of what the
world is like Natural phenomena do not always dovetail with such models,
however, and variations between experimental results and an idealized
conception of a particular relationship, behaviour, or procedure are common
The use of attribute hedges allows deviations between idealized models of
nature and instances of actual behaviour to be accurately expressed They
enable writers to restructure categories, define entities, and conceptualize
processes more exactly to distinguish how far results approximate to an
idealized state, specifying more precisely the attributes of the phenomena
described
In these examples, hedges are used to indicate variability with respect to
certain descriptive terms In (9), the degree to which the detected response is
considered 'normal' for the behaviour of those proteins is hedged, while in (10)
n is the comparison between two variables in terms of a given feature In (11)
'essentially' denotes a deviation from an ideal conception of a particular
process
(9) The response of the assembly of PSII proteins to the solute environment is unique
in some ways, but quite normal and predictable in others
(10) Although variable, the isoelectnc point of kumtz seed inhibitor is generally
lower
(11) The partially punfied PEPc kinase phosphorylates PEPc with a stoichiometry
approaching 1 and causes essentially the same reduction in sensitivity
 KENHYLAND 441

Attribute hedges indicate a discrepancy between actual results and either an

expected state or the concept routinely available to explain it, allowing a better
match with familiar descriptive terms Attribute hedges generally cluster around
this pragmatic core and involve the use of a finite set of items which Ernst
(1984) labels 'degree of precision' adverbs Such devices have received some
attention in the literature (eg Powell 1985, Channell 1990 and 1994)andare
similar to Prince et a/'s (1982) 'rounders' A number of these act as 'down-
toners" (Quirk et al 1972 509) which weaken the force of an attribute and can
be graded as to their strength, with those realizing the greatest effect almost
negating the force of the term modified (14)

(12) This shift could be partially caused by solvent-exposed a-helical segments of its
coiled-coil portion
(13) This appearance of kinase activity correlates quite well with
(14) while in the control, pTACC-C7 ACC synthase was barely detectable
Style disjuncts also function as attribute hedges as these indicate greater
precision in conveying the sense in which a proposition may be held to be true
(15) at an acidity that generally guarantees a quite stable assembly of the PS II
(16) decreases by approximately 60% at 44*c,
Finally, a qualification can indicate the precise standpoint from which to
judge the truth of a claim
(17) Viewed in this way the concept of lateral heterogeneity becomes obsolete
because the distinction between granal and stromal lamellae for
(18) Finally, from a practical point of view, the tRNAJip extended transcript could
serve to produce large quantities of wild type tRNAdSp transcripts
In each case, the hedge acts to qualify the predicate intensity or the validity of
the state of affairs expressed in the proposition In sum writers use attribute
hedges to seek precision in expression, and core examples encode variability,
rather than writer perspective

5 1 2 Reliability hedges
Reliability hedges indicate the writer's confidence in the truth of a proposition
They acknowledge subjective uncertainties and are motivated by the writer's
desire to explicitly convey an assessment of the reliability of propositional
validity They deal with the epistemically possible and contingent although such
subjective inferences can be confused with objective possibilities and often only
participant understandings can disambiguate a hedge from a verifiable
possibility Reliability hedges suggest the writer's reservations concerning
whether the situation actually obtains, keeping interpretations close to findings,
where claims may be less tenuous

(19) However, the opposite is also possible, and it cannot be ruled out that the 21 -kD
polypeptides seen in the bacteroid lane and in the soluble proteins lane are
442 HEDGING IN RESEARCH ARTICLES

(20) Together these results are consistent with the possibility, although do not prove
that the capacity to mediate the FR-HIR may be an intrinsic property of
phytochrome A

The principal motivation here is a desire to clarify the state of knowledge, a

hedge against complete accuracy, rather than protection against overstatement
In these core cases, acknowledgement of factual uncertainties predominates
over attempts to disguise the author's opinion
Reliability hedges are most commonly expressed by epistemic modal verbs,
epistemic adjectives, nouns, and adverbs These examples all seem to be largely
concerned with precision

(21) the high activity of NAIV raises the possibility thai the molecule of interest
may be IAA-valine
(22) Oscillations in fluorescence and O 2 evolution activity are probably an expression
of this short-term adjustment capacity
(23) it appears possible that the mechanism causing the light-activated fluores-
cence quenching may be triggered by either photosystem

Content disjuncts are commonly used to mark reliability as they comment on

the probability of propositional truth They include both adverbs of certainty,
which simply convey doubt on the information (24 and 25), or of mental
perception which show how results are understood (26)
(24) These screens were performed in white light, presumably precluding detection of
mutants in the FR-HIR
(25) This modification could possibly play a role in substrate binding
(26) Inactivation of the plastid genes is apparently not due to major recombinational
events at the level of ptDN A, since restriction enzyme patterns

As reliability hedges specify the actual state of knowledge, there may be some
hesitation about the strength of the connection between observed and assumed
events, as these two examples show
(27) Slow PS I turnover may favour QB binding to PS 11 or at least not prevent it
(28) This single mismatch may simply represent a laq polymerase error, or it may
result from amplification of a transcript from a different member of the PEPC
gene family
In both cases, confident conclusions are avoided by proposing alternative
formulations of a claim
All reliability hedges express a conviction about the truth of a statement as
warranted by the available facts, relying on inference, deduction, or repeated
experience Some examples, however, are difficult to distinguish from a
defensive lack of commitment While writer-oriented hedges similarly display a
reluctance to stray from the data, they carry an additional unwillingness to make
a commitment to conclusions Reliability hedges, on the other hand, essentially
make claims contingent due to knowledge limitations, and this is occasionally
made explicit
 KENHYLAND 443

(29) It is not known whether such a weak temperature response

(30) The mechanisms involved in the control of conductance through the stomata are
poorly understood The guard cells could respond to messages from the leaf
mesophyll, or have their own CO2 sensing mechanisms because they possess
chloroplasts
Jn summary, accuracy hedges contribute precision and work to specify a state
of knowledge rather than hedge the writer's commitment Both accuracy and
reliability types are principally concerned with interpretations of the world via
laws of reason and seek to increase the exactness of a claim, either by modifying
the sense in which terms describe reality or by stating a more precise appraisal
of certainty In either case, core examples focus on propositional precision The
writer's desire for protection from the possibility of error is either not implied,
as in attribute hedges, or is of secondary importance However, the line between
hedging confidence in the accuracy of statements and hedging commitment to
them is often blurred, and core cases, where either uncertainty or protection can
be exclusively distinguished, are relatively rare Contextual and formal con-
siderations can only identify the predominant function, rather than offer a
definitive categorization

5 2 Writer-oriented hedges
These limit the writer's commitment to statements Wnter-onented hedges
enable writers to refer to speculative possibilities while at the same time guard
against possible criticism, they are therefore often associated with higher-level
claims than accuracy-oriented ones Because the wnter is seeking to place
results in a wider context and demonstrate a contribution to the scientific pool
of knowledge rather than simply interpret findings, such claims carry a greater
risk and an element of self-protection may be necessary Simply, greater
generalization and interpretation require a greater degree of hedging (cf
Hunston 1994) Writer-onented hedges therefore create a clear pragmatic con-
trast with other content hedges accuracy-oriented hedges are proposition-
focused and seek to increase precision by refernng to the exact state of
knowledge or to how a proposition is to be understood, wnter-onented hedges
are writer-focused and aim to shield the wnter from the consequences of
opposition by limiting personal commitment These hedges thus diminish the
author's presence in the text rather than increase the precision of claims As
suggested by the fuzzy-set model, this is necessarily a matter of degree
While this usage follows Lakoff (1972) in associating hedges with 'fuzziness',
scientific wnters do not seek acceptance for claims through purposive
vagueness or fudging What is made fuzzy is the relationship between the wnter
and the proposition, rather than the claim itself These hedges help minimize the
scientist's personal involvement and thereby reduce the probability of
refutation This allows wnters to 'anticipate and discountenance negative
reactions to the knowledge claims being advanced' (Swales 1990 175) Pnnce
et al (1982), Rounds (1982), Powell (1985), and Nash (1990) have all
suggested that hedges may serve as an insurance in helping wnters protect their
444 HEDGING IN RESEARCH ARTICLES

reputations and limit the damage which may be incurred from categoncal
commitments
In core examples, care is taken to avoid assuming explicit responsibility for an
assertion while seeking to secure 'uptake' by moving the reader to the writer's
standpoint
(31) The present work indicates that the aromatic nng to which the carboxyl group is
bound is not necessary, provided that a bulky substituent is present
(32) Although it is premature to answer this question it might be suggested that
synthetases present in nuclei [33] could be involved in the regulation of the
processing of
The most distinctive signal of wnter-onented hedges is the absence of writer
agency The author's responsibility can be reduced by the use of passive
constructions (33), clausal subjects (34), or the construction of'abstract rhetors'
which nominahze a personal projection (35-37)

(33) The BS fraction is assumed to originate from the center of the

(34) It might be speculated that the lack of crDNA methylation in cv Platenese could

(35) These data indicate that phytochrome A possesses the intrinsic

By foregrounding 'These data' (35), the writer presents a view where data,
vested with agentivity, are attnbuted with primary responsibility for an
interpretation, they become the source of the claim
Judgemental epistemic verbs, particularly speculative (e g assume, predict,
propose) and evidential verbs (e g appear, seem), in impersonal phrasings are a
principal means of withholding personal commitment The tentativeness relates
mainly to the commitment the author wishes to bestow on the statement rather
than a stnct concern with the truth of its propositional relationships
(36) Circumstantial evidence indicates specific shifts in the mRNA levels for
(37) The theoretical Ime assumes that the partial pressure of CO2 is the same in the
(38) The model implies that the function of granna is to shield
(39) From this discussion, then, it would appear that some of the changes in amino
acid concentrations
Another strategy to achieve distance from a claim is to attribute its source, or
underpinnings, elsewhere Thus, writers may refer to wider bodies of knowledge
when moving away from what can be confidently implied by their results or
methods
(40) This hypothesis seems plausible because UV-B-absorbing flavonoids
accumulate in leaf epidermal cells, where they may protect the inner cell layers
from UV-B damage (CaldweHetal 1983, Beggsetal 1986)
Wnter-onented hedges often accent the procedures and methods of science, a
feature less salient in other forms of hedging Indeed, reference to methods, the
models employed, or the conditions under which the results were obtained are
an important means of hedging personal commitment
 KENHYLAND 445

(41) Under these conditions phosphorylation of PEPc by the partially purified kinase
caused a concomitant 3- to 4-fold increase in the
(42) The prediction of this model is that aminooxyiovalerates with aromatic rings
characteristic of potent synthetic auxins would be good candidates for active
molecules
(43) Despite the limitations of this method, the results suggest that the protein
mentioned
In sum, these devices function to hedge the writer's commitment to preposi-
tional content while leaving its truth value open The avoidance of a strong
alignment with a proposition is the defining feature of writer-oriented hedges
They allow authors to seek acceptance for the highest-level claim they can for
their results while protecting them from the full effects of its eventual overthrow
Indeed, a hedge can alter the context of the statement in which it occurs so that
the resultant claim is very hard to falsify Writer-oriented hedges diminish
personal responsibility for a variety of reasons (e g Latour and Woolgar 1979
75-87), but they are often the result of the preliminary character of much scien-
tific research, which accordingly has to be hedged against later falsification
Often results are presented in areas where hypothesis testing is extremely diffi-
cult, but the forces of discovery and recognition provide a context in which
productivity is a measure of professional worth and which results in the need for
early and frequent publication

5 3 Summary of content-oriented hedges

Overall, content-oriented hedges are a major communicative resource as they
enable scientific wnters to both negotiate the precision of claims and convey an
attitude to them On one hand, the writer seeks to present statements with
appropriate accuracy, on the other, to make the strongest claim possible while
limiting the damage of error This is principally achieved by a repertoire of
lexical markers and various hedging strategies, although it is difficult to
exclusively assign specific devices to particular functional categories. Fre-
quently, statements include several functions, thus in (44) the presence of an
abstract rhetor, evidence, and supporting testimony indicates wnter-
onentation Reference to the limited state of knowledge and alternative read-
ings carries reliability functions
(44) Thus the evidence strongly favours the conclusion that phytochrome A is the
primary, if not exclusive, FR-HIR photo receptor, consistent with previous
physiological studies (Beggs et al 1980, Holmes & Schaffer, 1981, Smith &
Whiteiam, 1990) On the other hand, the possible involvement of phytochromes
C, D, or E in an interactive or synergjstic manner with phytochrome A cannot
presently be ruled out

The blurred distinction between qualification of confidence and of commitment

means particular forms will invariably carry elements of both meanings In
terms of a fuzzy-set model, core examples of content-oriented hedges will
446 HEDGING IN RESEARCH ARTICLES

contain proposition-focused meanings, but the majority of cases are likely to

occur at the periphery and convey shades of wnter-focused meanings

6 READER-ORIENTED HEDGES
While hedging is traditionally linked with the objective dimension, securing
ratification of scientific claims also involves reducing the risk of negation on
subjective grounds Core examples of reader-onented hedges confirm the
attention writers give to the interactional effects of their statements

6 1 Motivation for reader-onented hedges

It was argued above that the functional similarities of lllocutionary mitigation in
conversation and research writing suggested by Myers (1989) neglects
important differences between the two domains A straightforward application
of imposition, distance, and social power cannot fully explain scientific hedging
I suggest that collegiate deference occurs both as a response to discourse norms
of communication and the fact apodictic statements are inherently face
threatening to peers
Essentially, in presenting claims a writer also projects a persona which carries
information concerning the writer's professional attitudes to the discipline (cf
Campbell 1975) Tins professional personality is crucial to achieving rhetorical
goals as it also conveys an attitude about the reader and his/her role in the
negotiation of knowledge claims Presenting claims as ex-cathedra assertions
displays an unacceptable deviant persona as it ignores any involvement by the
reader in the ratification of claims Categorical assertions leave no room for
negotiation they imply an assurance in the certainty of arguments that require
no feedback, and this relegates readers to a passive role Hedged statements, on
the other hand, mark claims as provisional, they invite the reader to participate
in a dialogue Hedges solicit collusion by addressing the reader as an intelligent
colleague capable of participating in the discourse with an open mind Good
arguments are only 'good' from a particular perspective and hedges work to
create this perspective Once this is achieved, arguments can be based on other
criteria
Together with this interpersonal dimension, there is a normative aspect of
reader-onented hedges Acting as a scientist involves implicit and informal
"rules'of communicative behaviour (eg Becher 1989) These rules also govern
scientists' communication styles, where expectations include an obligation on
the writer to defer to the views of colleagues, adhere to limits on self-assurance
and engage in debate with peers So while a paper must carry conviction, it must
also appeal to the reader as a knowledgeable scientist, able to decide about the
issues presented
Convention, rather than calculation of personal cost, may drive this system,
but it is underpinned by the possibility of sanctions which can influence one s
career These operate through the peer validation of claims The fact that
knowledge accreditation is a social process means the achievements and
reputation of every scientist are in the hands of colleagues Having research
 KEN HYLAND 447

contributions accepted in reputable journals being cited by others, acquiring

research grants, receiving professional honours, etc are badges of success
which depend on adherence to scientific norms The mere act of participating in
the formal communication system emphasizes the extent to which the
individual's investigations and discoveries are dependent on the work and
acceptance of others Publishing a paper involves entering into an interactional
contract with one's readers concerning the conduct of scientific debate, the
elaborate communication system which manages publication and validates
knowledge thus also regulates authors and helps enforce stylistic conventions
This framework of institutional obligations enables participants to get on with
the task at hand

6 2 Forms of reader-oriented hedges

Again, the data does not allow functions to be unequivocally distinguished since
many devices are used to convey both reader- and writer-based hedges Core
cases of reader-oriented hedges are most obvious, however, in managing
substantive disagreement and avoiding conflict

(45) We do not know the reason for the discrepancy between our results and those of
Ngernpraintsin et al, but it might reflect genetic differences in the cultivars
employed
(46) Quick et al also noted that there was no consistent effect of reduced Rubisco ( )
Our results do not support the latter observation, because conductance
decreased with increasing p, in both control and anti-SSu plants {fig 5)
(47) There has been some disagreement concerning whether illumination increases
the Vma, of PEPc (eg refs 1,2,4,11) In our hands, there was no significant change
in Vnun
mj. on illumination

A vanety of devices is used here to soften the effect of criticism While modal
auxiliaries and an admission to a lack of knowledge express an uncertainty
which avoids direct criticism, it is the choice of personal subjects which is
cntical The frequency of first person pronouns (whether singular, or, in the case
of multi-authored papers, plural, as here) in science is confirmed by Banks
(1993) who found they occurred four times more often with hedging verbs than
other verbs But while Banks believes this serves to 'trim' or reduce the force of
the hedge, my interpretation is that an overt acceptance of personal responsibil-
ity mitigates the expression of a proposition and signifies a reader-oriented
hedge
As Myers points out, expressions of personal belief weaken claims because
they are inconsistent with the supposed universality of scientific knowledge
(Myers 1989 14) Reference to the writer's direct involvement in the research is
therefore a conscious strategy to subtly hedge thegenerahzabihty of a claim and
mark a position as an individual interpretation Reference to the wnter
explicitly marks a statement as an alternative view rather than as a definitive
truth, the hedge signals a personal opinion, allowing the reader to choose the
more persuasive explanation The use of impersonal expressions to comment
448 HEDGING IN RESEARCH ARTICLES

on others' work, on the other hand, suggests a writer-oriented hedge Where

writers are more concerned with protecting their credibility than with the non-
ratmcation of their argument criticisms are often more diffuse, with a tendency
to avoid attributing what is criticized to a particular source
(48) Thus the existence of multiple gene families and expression of multiple ACC
synthase isoforms question the validity of earlier results obtained with either
crude plant extracts or partially purified enzyme preparations
(49) In spite of its shortcomings, the method has been widely employed to evidence
this type of modification in a number of genomes including plastid and nuclear
ones

In addition to weakening criticism, personal attribution is also used to soften

claims This makes frequent use of epistemic verbs, particularly verbs of
judgement (50 and 51) and deduction (52 to 54)

(50) Thus we propose that this insert is the major site of interaction with the
membrane
(51) I believe that the major organisational principle of thylakoids is that of continu-
ous unstacking and restacking of sections of the membrane
(52) Of this percentage, we calculate that more than halt is present in the coiled-coil
portion and the rest in the globular domain
(53) We infer that the rate becomes limited by the rate of regeneration of RuBP-
saturated Rubisco kinetics,
(54) Our interpretation of these results is that the total level UV-B

Thus, by specifying a personal source, t h e writer leaves the claim open to the
reader's judgement, often combining with a writer-oriented hedge to specify the
limits within which the claim holds

(55) Under these conditions, we predicted that seedlings would respond through the
action of phytochrome B, and thus allow

Deference to the reader may also be achieved by offering a claim as one possibil-
ity among many, using the indefinite article (56) or hypothetical conditionals
(57) to suggest alternatives
(56) a model implying almost complete lateral heterogeneity in the thylakoid
membrane
(57) If we assume that the apparent molecular weight obtained by SDS PAGE is
correct, this suggests that only a few ammo acids are missing from the N-terminal
end and that a leader sequence may be encoded

Questions may also serve as hedges These appeal to the reader as a fellow
scientist involved in the problems of empirical research and engaged in the
communahty of the scientific quest

(58) Could such a putative interaction of an aminoacyl-tRNA synthetase with

precursor tRNA have a physiological significance9
(59) How is it, then, that this gene from a species that never forms nodules has the
capacity to be expressed at elevated levels m nodules9
 KENHYLAND 449

The writer thereby hedges the claim to be made by explicitly drawing the reader
into the deductive process, rhetorically treating the audience as capable of
making the same logical inferences
In sum, these hedges recognize the need for reader acceptance in accrediting
knowledge and respond to the possibility of opposition to claims on inter-
personal grounds Here, writers consider both the reader's role in confirming
knowledge and the need to conform to community expectations regarding
deference to colleagues' views Core examples are therefore distinguished by
features addressed to the needs of an audience which anticipates involvement in
negotiating claims

7 FUNCTION AND FORM IN SCIENTIFIC HEDGING

Scientific writers are onented both to what they say and who they are saying it to
and the type of hedge used conveys a choice in how to best negotiate the ratifica-
tion of their claims Three main functions of hedging in scientific RAs have been
suggested First, to present claims with greater precision with respect to both the
terms used to describe real-world phenomena and the degree of reliability the
writer invests in the statement Second, to signal reservations in the truth of a
claim to limit the professional damage which might result from bald proposi-
tions Third, to give deference and recognition to the reader and avoid
unacceptable over-confidence The first two relate to the strictly epistemic func-
tions discussed by Lyons (1977), Palmer (1990), and others, and express doubt
in statements The third concerns the writer's contribution to the development
of a writer-reader relationship in gaining reader ratification These functions
are summarized in Table 1
Because it is writers, not sentences, that hedge, contextual understandings
play a crucial role in the design and interpretation of scientific arguments In
particular, the confirmation of claims and the rewards of publication help clarify
the use of hedges in science Essentially, writers seek agreement for the
strongest claims they can for their evidence while neutralizing the possibility of
opposition by meeting both adequacy conditions, between the proposition and
the world, and appropnacy conditions, between the proposition and the reader
Whether a writer will choose to hedge or present information categorically
therefore depends on an assessment of the non-linguistic context, the extent to
which the proposition corresponds with what is believed to be true of the world
and its potential for eliciting a required reader response
However, a principal feature of hedging is indeterminacy The maximally
efficient exchange of information in science means that particular forms are
always likely to carry more than one meaning Any classification scheme
inevitably distorts reality by imposing hard and fast categories on the fluidity
and indefiniteness of natural language use, but a fuzzy-set model captures this
feature by enabling salient aspects of membership to be identified while
conceding that individual cases are likely to express several meanings simul-
taneously These observations allow us to identify the devices most often
associated with the expression of core cases (Table 2)
450 HEDGING IN RESEARCH ARTICLES

Table I Principal functions of statements in scientific writing

T>pe Function Main hedging effect of core examples

content commitment assertiveness

Categorical assertion full commitment to p and truth

value
Reader-oriented hedges expression to gain reader
acceptance of p
Content-oriented hedges the extent to which p
accounts for phenomenon
Accuracy-onented hedges correspondence of p to
reality
Wnter-onented hedges writer commitment to p

*/ = presence of a feature
X — absence of a feature

Table 2 Summary of hedging functions and principal realization devices

Content-oriented Reader-oriented

Accuracy-oriented Writcr-oneiited

Hedges propos111on.1l con lent Hedges writer commitment Hedges assertiveni.\s

Attribute type Epistemic lexical verbs Epistemic lexical verbs
Precision adverbs judgemental judgemental
content disjuncts evidential deductive
style disjuncls Impersonal expressions Personal attribution
downioners passive voice Personal reference to
Relidhihly type abstract rhciors methods
Cpistemic lexical verbs emply subjects model
Epistemic modal adjectives Thematie epistemic device Olfer alternatives
Epistemic modal nouns Attribution lo literature conditionals
Content disjunct adverbs Impersonal reference to indefinite articles
Limited knowledge method Involve reader
model direct questions
experimental conditions reference to testability
Assumption of shared goals
Hypothetical e g would

8 HEDGING IN SCIENCE AND IN OTHER DOMAINS

A further important question is whether these functions are specific to scientific
RAs While a comparative analysis was not undertaken, it is uncertain whether
this would reveal conclusive evidence of distinctiveness, as particular varieties
 KENHYLAND 451

or genres may not actually be characterized by particular language conventions

(e g Paltndge 1993) However, some tentative conclusions can be drawn based
on those domains which have received attention in the literature
Hedging has received the most attention in casual conversation where it is
perhaps twice as frequent as in written discourse and represents a significant
resource for speakers (e g Coates 1987, Holmes 1995) Conversationalists, like
scientists, employ hedges to convey both affective and referential meanings, but
there are important differences in these functions While it is apparently
common in speech, for example (eg Powell 1985, Channell 1994), scientific
writing exhibits little use of deliberate vagueness in the sense of withholding
necessary information Vagueness is simply not a viable communicative option
in most scientific writing where propositional precision is central to persuasive
argument as it is the only control a writer has in influencing the reader's views
Nor do writers seek to create conviviality or acknowledge power differences,
and while hedges in scientific writing involve sensitivity to addressees, this
differs considerably from conversational goals of negative politeness or phatic
communion
Functional differences also exist between science RAs and other spoken
genres Dubois (1987), tor example, found hedges in bio-medical conference
presentations served to deliberately diminish precision in order to hedge
accuracy, foreground more exact figures, and cite theories believed to be in
error These functions play a very insignificant role in molecular biology RAs,
however, and again, it is doubtful whether the genre encourages such uses
Similarly, the use of hedges in doctors' discussions to explicitly indicate that
more precise terms are not relevant (Prince et al 1982) was not observed in my
data In addition, the use of hedges to help structure organizational roles and
negotiate discourse statuses between participants in academic counselling
encounters (He 1993) does not occur in the RAs Writer-reader relationships
in science RAs are institutionally and normatively defined by conventions of
interpersonal conduct and underlined by community sanctions Hedges may
therefore help to project a persona and demonstrate adherence to the 'rules' of
communication, but this is simply to achieve higher-level functions of gaining
acceptance of knowledge claims, rather than to structure interpersonal relation-
ships per se
Several written genres also contrast with biology RAs in the use of hedges
Textbooks, for example, reserve hedges for matters which lack a consensus
(Myers 1992), while a taxonomy of hedges in medical case reports and research
papers (Salager-Meyer 1994) admits a very reduced role for the expression of
doubt compared with the scheme described here The role of hedges to convey
uncertainty in medical RAs has, however, been confirmed by Adams Smith
(1984)
Nor is the rhetorical manipulation of hedges to present uncontroversial
material and deflect criticism from more contentious areas, found in the literary
criticism of F R Leavis (for example, Simpson 1990), appropriate in scientific
contexts In science, readers place interpretive reliance on the use of hedges to
452 HEDGING IN RESEARCH ARTICLES
distinguish the credibility of claims (reliability hedges) and the extent of writer
commitment to them (writer-oriented hedges) Similarly, the use of hedges
found in economics RAs (Channell 1990) to support claims by deliberately
fudging accurate numerical information, without violating the conventions of
truthfulness, is not a feature of my data However, ChannelPs observations con-
cerning the need to balance persuasive writing and factual reporting certainly
apply in science The journal corpus suggests that while data are presented as
precisely as required for the practical purposes of convincing peers of a claim,
expectations of exactitude may be higher in this discipline
Thus, while hedging is found to be important in a variety of domains, the
circumstances recognized as appropriate for its use and the functions it is seen
to fulfil often differ markedly Confirmation of these differences, however,
awaits further research

9 CONCLUSIONS
I have argued that hedges in scientific texts are the result of informational,
rhetorical, and personal choices which cannot be fully understood in isolation
from social and institutional contexts Linguistic analyses alone cannot provide
a rationale for such choices and the framework proposed here seeks to reflect
this interpretive environment Research articles clearly reveal the relationship
between a discourse community, standards of knowledge, and textual
representations, and it is these in combination which motivate the use of hedges
Hedges are abundant in science and play a critical role in academic writing
more generally They constitute an essential element of argumentation in
presenting new claims for ratification, and are among the primary features
which shape the research article as the principal vehicle for new knowledge An
understanding of their use therefore has important implications for a number of
areas, and can contribute to the growing literature on the rhetoric of science,
revealing important insights into how science establishes its claims to
knowledge and how scientists carry out their work Information about hedging
can also advance our understanding of the practice of evidential reasoning and
also has practical consequences in ESP where textbooks often emphasize the
impersonality of scientific discourse and either ignore hedges or advise students
to avoid them completely (Hyland 1994) Most importantly, however, the
analysis demonstrates the dynamic and interactive nature of scientific writing It
contributes to a growing sociological and linguistic interest in professional writ-
ing by providing a discourse analytic understanding of one means by which
scientific discourse is both socially situated and structured to accomplish
rhetorical objectives
(Revised version received November 1995)
 KENHYLAND 453

RErERENCES
Adams-Smith, D. 1984 Medical discourse Aspects of authors comment' English for
Specific Purposes 3 25-36
Banks, D. 1993 "Hedges and how to trim them' in M Brekke O Anderson, T Dahle,
andJ Mykmg(eds) 1993 9th European symposium of LSP University of Bergen
Bazerman, C 1985 "Physicists reading physics Schema-laden purposes and purpose-
laden schema* Written Communication 2/\ 3-23
Becher, T 1989 Academic Tribes and Territories Intellectual Inquiry and the Cultures
of Disciplines Milton Keynes SRHE/Open University Press
Brown,P a n d S Levinson 1987 Politeness Some Umversals in Language Usage Cam-
bridge Cambridge University Press
Brufee, K 1986 Social construction Language and the authority of knowledge A
bibliographical essay * College English 48 772-9
Butler, C 1990 Qualifications in science Modal meanings in scientific texts* in W Nash
<ed) 1990b
Campbell, P 1975 The personae of scientific discourse' Quarterly Journal of Speech
61 391-405
Channell, J 1990 "Precise and vague expressions in writing on economics in W Nash
(ed) 1990b
Channell, J 1994 Vague Language Oxford Oxford University Press
Coates, J 1983 The Semantics of the Modal Auxiliaries Beckenham Croom Helm
Coates, J 1987 Epistemic modality and spoken discourse Transactions of the
Philological Society 85 100-31
Cnsmore, A , R Markkaned, and M. Steflensen 1993 Metadiscourse in persuasive
writing A study of texts written by American and Finnish university students * Written
Communication 10/1 39-71
Dubois, B 1987 "Something on the order of around forty to forty four' Imprecise
numerical expressions in biomedical slide talks' Language and Society 16 527-41
Ernst, T B 1984 I awards an Integrated Theory of Adverb Position m English Indiana
Linguistics Club
Fahnestock, J. 1986 'Accommodating science The rhetorical life of scientific facts
Written Communication 3/3 275-96
Faust, D 1984 I he Limits of Scientific Reasoning Minneapolis University of
Minnesota Press
Hanania, E and K. Akhtar 1985 "Verb form and rhetorical function in science writing
A study of MS theses in biology, chemistry and physics' ESPJournal 4 49-58
He, A. 1993 "Exploring modality in institutional interactions Cases from academic
counselling encounters' Text 13/2 503-28
Holmes, J- 1984 'Modifying illocutionary force * Journal of Pragmatics 8 345-65
Holmes, J. 1995 Women, Men and Politeness London Longman
Hunston, S 1994 "Evaluation and organization in a sample of written academic
discourse* in M Coulthard (ed) 1994 Advances in Written Text Analysis London
Routledge
Hyland, K 1994 "Hedging in academic wnting and EAP textbooks' English for Specific
Purposes 13/3 239-56
Hyland, K. 1996 Talking to the academy Forms of hedging in science research articles*
Written Communication 13/2 251-81
454 HEDGING IN RESEARCH ARTICLES
Kourany, J. A. 1987 Scientific Knowledge Basic Issues in the Philosophy of Science
Belmont, CA Wadsworth
Kuhn,T 1970 The Structure ofScientific Revolutions (2ndedn) Chicago Universityof
Chicago Press
Lakoff, G 1972 'Hedges A study in meaning cntena and the logic of fuzzy concepts'
Chicago Linguistic Society Papers 8 183-228
Latour, B. and S. Woolgar 1979 Laboratory Life I he Social Construction of Scientific
Facts Beverly Hills Sage
Longino, H. 1990 Science as Social Knowledge Values and Objectivity in Scientific
Inquiry New Jersey Princeton University Press
Lyons, J. 1977 Semantics Vols I and 2 Cambridge Cambridge University Press
Myers, G. 1985 The social construction of two biologists' proposals' Written
Communication 2/3 219-45
Myers, G 1989 "The pragmatics of politeness in scientific articles * Applied Linguistics
10/1 1-35
Myers, G. 1992 "Textbooks and the sociology of scientific knowledge' English for
Specific Purposes 11 3-17
Nash,W 1990a "Introduction The stuff these people write' in W Nash (ed ) 1990b
Nash, W (ed) 1990b The Writing Scholar Studies in Academic Discourse Newbury
Park, CA Sage
Palmer, F 1990 Modality and the English Modals (2ndedn) London Longman
Paltridge, B 1993 'Writing up research A systemic functional perspective 'System 21/
2 175-92
Perfetti, C A and D McCutchen. 1987 'Schooled language competence Linguistic
abilities in reading and writing' in S Rosenberg (ed) 1987 Advances in Applied
Psycholinguisttcs Vol2 Cambridge Cambridge University Press
Pindi, M and T Bloor 1986 "Playing safe with predictions Hedging, attribution, and
conditions in economic forecasting' Written Language BAAL 2 CILT
Powell, M 1985 'Purposive vagueness An evaluation dimension of vague quantifying
expressions' Journal of Linguistics 21 31 -50
Prince, E.,J Frader, and C. Bosk 1982 On hedging in physician-physician discourse
in R D Pietro(ed) 1982 Linguistics and the Professions Hilisdale, NJ Ablex
Quirk, R, S Greenbaum, G Leech, and J Svartvik 1972 A Grammar of
Contemporary English Harlow Longman
Rounds, P 1982 Hedging in Written Academic Discourse Precision and Flexibility The
Universityof Michigan Mimeo
Salager-Mejer, F 1994 'Hedges and textual communicative function in medical English
written discourse' English for Specific Purposes 13/2 149-70
SCI. 1993 1992 Journal Citation Reports (JCR) Philadelphia Public Institute for
Scientific Information
Simpson, P 1990 "Modalityin literary-cnucaldiscourse inW Nash (ed) 1990b
Skelton, J 1988 'Comments in academic articles'in P Grunwall(ed) 1988 Applied
Linguistics in Society London C1LT/BAAL
Swales, J. 1990 Genre Analysis English in Academic and Research Settings
Cambridge Cambndge University Press
Van Frasen, B 1980 The Scientific Image Oxford Oxford University Press
Vande Kopple, W. 1985 "Some exploratory discourse on metadiscourse College
Composition and Communication 36 82-93
Zadch, L A 1972 'A fuzzy-set-theoretic interpretation of linguistic hedges Journal of
Cybernetics 2/3 4-34