You are on page 1of 8

Defining and responding to plagiarism 33

Elizabeth Wager
Defining and responding to plagiarism

Learned Publishing, 27: 33–42


Defining and
responding to
Plagiarism has caused problems for editors
and publishers for centuries. The origin of the
word may be traced back to the Roman poet
Martial (c. AD 80), who claimed that another
poet recited ‘my books to the crowd as if none Elizabeth WAGER
other than your own’. The term was probably Sideview
first used in English, in its current sense of lit-
erary theft, sometime in the 15th century.1 In ABSTRACT. A clear definition of plagiarism and
the 21st century, awareness of, and interest in, the ability to classify it into more or less serious forms
would help editors and publishers to devise policies to
plagiarism has been fuelled by the availability
handle this problem. Text-matching software is a useful
of text-matching software which can reveal the tool for measuring the extent of text copying but it
use of ‘copy–paste’ by authors. However, while cannot detect plagiarized tables or figures, plagiarism
software (such as iThenticate, eTBlast, or use of ideas, or plagiarism in translation. Furthermore, a
of search engines) makes it easy to detect text working definition of plagiarism in relation to research
matches, not all such matches are necessar- reports needs to take into account factors such as the
ily plagiarism. Text matches may occur for originality of the copied material, its position in the
legitimate reasons such as: co-publication of report, the adequacy of referencing, and the intention
guidelines (agreed by several journals), repub- of the authors as well as the extent of the copying.
lication with the agreement of original and This article considers all these factors and proposes
new publishers to make documents available possible definitions of major and minor plagiarism in
to new audiences (as happened, for example, relation to scholarly publications which might be used
with this article), online availability of confer- as the basis for anti-plagiarism policies in conjunction
ence abstracts resembling later journal publi- with resources such as the COPE flowcharts.
cations, or the existence of several versions of
a document (e.g. in preprint servers or institu-
tional repositories as well as journal websites).
Reuse of short word strings may be accidental
or unavoidable, and exact repetition of tech-
nical descriptions (e.g. in a methods section)
may be helpful to ensure accuracy. Therefore,
editors and publishers need workable defi-
nitions of plagiarism and probably a more
detailed taxonomy to distinguish the different
forms of plagiarism than that provided by dic-
tionary definitions.
Once they have defined plagiarism, and
armed with sophisticated tools for detecting it,
editors need to decide how to handle it.
© Elizabeth Wager 2014
Note: this article is adapted from the COPE discussion
Background/existing guidelines document ‘How should editors respond to plagiarism?’
which was first published at
The COPE flowcharts2 recognize that an edi- Discussion%20document.pdf in April 2011 under a Creative
tor’s response to plagiarism should depend on Commons Attribution License. Elizabeth Wager


Defining and responding to plagiarism 35

the type and extent of the copying. They sug- • referencing/attribution

gest different responses to ‘Clear plagiarism’ • intention
(described as ‘unattributed use of large por- • author seniority
tions of text and/or data, presented as if they • language.
were by the plagiarist’) and ‘Minor copying of
short phrases only’ with ‘no misattribution of Extent
data’ (giving an example of copying ‘in [the] The most blatant forms of plagiarism involve
discussion of [a] research paper from [a] non- the copying of entire papers or chapters which
native language speaker’). The flowcharts also are republished as the work of the plagiarist.
distinguish plagiarism (i.e. copying from oth- Such cases usually involve not only plagiarism
ers) from redundancy or ‘self-plagiarism’ (i.e. but also breach of copyright. Whole articles or it is important
copying from one’s own work). The flowcharts chapters may also be plagiarized in translation.
also suggest that the editor’s response might to recognize
The COPE retraction guidelines recommend
vary according to the seniority of the author that such articles should be retracted, and that ideas,
(with editors simply writing an educational the flowcharts on plagiarism suggest that edi- images, creative
letter to very junior researchers but consider- tors should consider contacting the author’s
ing informing the institution of more senior
works (e.g.
institution in such cases.3 However, the COPE
authors) as well as whether the authors are retraction guidelines state that:
musical com-
writing in their native language. positions
[I]f only a small section of an article (e.g. a
few sentences in the discussion) is plagiarised, or choreo-
Types of plagiarism editors should consider whether readers graphies), and
(and the plagiarized author) would be best data can be
Any original creation may be plagiarized. served by a correction (which could note the
Although most discussions focus on text fact that text was used without appropriate plagiarized
(and this type of copying is usually the only acknowledgement) rather than retracting the
kind that can be detected using software) it entire article which may contain sound, original
is important to recognize that ideas, images, data in other parts.
creative works (e.g. musical compositions or
choreographies), and data can be plagiarized. Scholarly works often summarize the work of
The following factors may be helpful in dis- other researchers. It may be difficult to draw a
tinguishing types of plagiarism (see Table 1): line between legitimate (and accurate) repre-
sentation of other studies and copying original
• extent material. Researchers may also feel that little
• originality of copied material harm is done if they use similar language to
• position/context another publication so long as the source is

Table 1. Features of different types of plagiarism

Feature Least severe type Most severe type
Extent A few words A few Whole Several Whole paper
sentences paragraph paragraphs
Originality of copied Widely used Phrase/idea Original phrase/
material phrase/idea used by a small idea
number of
Position/context/type of Standard Describing Data/findings
material method another worker’s
Referencing/attribution Source fully Source partially/ Unreferenced
and clearly inaccurately
referenced referenced
Intention No intention to Intention to
deceive deceive


36 Defining and responding to plagiarism

Table 2. Examples of language of low originality used in reports of medical research

Phrase Hits for exact phrase (January 2011)
Google Google Scholara
P < 0.05 was considered statistically significant 588,000 70,600
performed according to the Declaration of Helsinki 410,000 1,860
double-blind, double-dummy, placebo-controlled 56,800 882
numbered, opaque, sealed envelopes 12,200 912
randomized in a 1:1 ratio 8,510 1,020
computer-generated random number list 5,120 354

Google Scholar searches for academic publications only.

properly cited. If the original authors summa- original to be attributed to the creator). While
rized their findings clearly and succinctly it publishers of poems and song lyrics tend to
could be argued that little is gained by forc- guard their copyright fiercely, and permission
ing other authors to paraphrase this. However, is required to quote even a single line, tech-
others will argue that any verbatim copy- nical publications may contain descriptions of
ing should be indicated by using quotation standard techniques which are described in
marks, otherwise they would consider it to be similar or identical ways because the language
plagiarism. is not considered original. Therefore, the orig-
Most text-matching software detects strings inality of the copied material should be con-
extent alone of several words, since duplication of just sidered as well as the extent.
cannot be taken a few words can occur by chance. However,
as a benchmark academic papers and reports may contain Position/context
technical language that involves standard Certain sections of research reports may be
phrases that are longer than the strings used more likely to include non-original material. In
by software. For example, a Google search for particular, the methods section may describe
the phrase ‘smokers with chronic obstructive widely used techniques. The use of standard-
pulmonary disease’ produces >58,000 results, ized descriptions of public data sources, pro-
suggesting that this is a widely used phrase, prietary techniques, questionnaires, or equip-
but such a six-word string may also trigger a ment may even be regarded as helpful to
match on text-matching software. Therefore ensure accuracy and consistency. For example,
extent alone cannot be taken as a benchmark. analysis of the UK General Practice Research
Table 2 shows the frequency of some phrases Database has resulted in over 750 publica-
commonly seen in medical publications which tions. All these publications probably include
are so widely used their use would not be con- a description of the database and these are
sidered as plagiarism but which might count as likely to use similar language. Similarly, it
matches when using text-matching software. may be better if the original description of an
assay provided by a company or the supplying
Originality of copied material laboratory is copied rather than reworded by
Originality needs to be considered in con- each user, since the original wording may be
junction with extent. The example given the most accurate. Therefore editors may view
above indicates the difference between a stan- text similarity in methods sections as unavoid-
dard, technical phrase (such as ‘smokers with able or even desirable and therefore treat it
chronic obstructive pulmonary disease’) and differently from similarity occurring in other
original usage (such as Winston Churchill’s parts of a paper.
‘the end of the beginning’ or Shakespeare’s The editor of a mathematics journal has
‘the winter of our discontent’ – both of which noted:
contain fewer than six words and would there- [S]tatements of the mean value theorem from
fore probably not be detected by text-matching calculus book to calculus book are virtually
software yet are usually considered sufficiently identical; there’s really only way to state


Defining and responding to plagiarism 37

Schur’s Lemma. Probably, automated software will, naturally, contain large sections from the
would detect these instances as plagiarism. previous versions and this may appear to be
They’re not, of course. Sometimes, there’s plagiarism if the authors have changed (since
really only one way to define something or give automatic systems will not recognize acknowl-
the ‘usual examples’. (Lance Small, personal edgements to previous versions).
The type of publication may also affect judge- Referencing/attribution
ments about the acceptability of text simi- Academic publications are expected to refer-
larity. While research reports may describe ence other works and may also quote from
standard methods, editorials may be expected them. Inexperienced or poorly trained authors
to represent the author’s opinion and origi- may mistakenly believe that so long as another
nal views on a topic and it would therefore work has been cited, parts of it can be repro-
be considered inappropriate to use the same duced in their own work. While copying parts
words as another author except in direct of cited work is probably not intended to
and properly attributed quotations. Similarly,
review articles, and the discussion sections of
deceive the reader in the same way as copy- inexperienced
ing unattributed material, the practice is gen- or poorly
research papers, are expected to provide an erally considered to be poor scholarship and
original synthesis of, and commentary on, pre- inappropriate for an academic journal. Editors trained authors
viously published work. Therefore, apart from may have a role in educating authors if they may mistakenly
quotations, the words may be expected to be discover this type of copying, especially if it is
the author’s own. believe that so
detected before publication.
However, editors may also consider the long as another
consequences of the copying and its potential Intention work has been
to mislead readers. In this respect, copying a
few sentences from the discussion section of Intention to deceive is often considered a cited, parts
another researcher’s paper may be consid- factor distinguishing misconduct from care- of it can be
ered less harmful, and less deceitful, than the less work or honest error. However, it is usu- reproduced in
theft of data (which may constitute not only ally impossible to prove intent and therefore
may be less useful in practice than in theory.
their own work
plagiarism but also data fabrication since the
work being reported was not actually done Extreme forms of plagiarism, such as copy-
by the copier). Thus, if a submitted article ing an entire paper and submitting it under
that appears to describe legitimate, original a different author’s name to another journal,
research, includes some sentences taken from can only be deliberate. Editors must use their
the discussion of another author’s paper on a own judgement to determine whether authors’
related topic, the editor may simply ask the explanations for less extreme forms of copying
author to indicate that these are direct quota- are plausible or could have occurred through
tions, or to paraphrase the copied text, before honest error. When confronted with identi-
publication. If the copying is discovered after cal text, authors may counter with explana-
publication, the editor may suggest that it tions such as having a photographic memory
can be rectified by a correction rather than a or inadvertently copying notes or preliminary
retraction and may not feel that the author’s material into a publication. When a senior
institution should be informed. researcher at Stanford University was found
When using software to detect text simi- to have incorporated large chunks of text
larity, editors should not forget that refer- from a well-known textbook into a chapter
ence sections will contain large amounts he had prepared for another book, he told
of copied text in the titles of cited articles. an inquiry that ‘when he cut-and-pasted the
Some software systems, such as iThenticate/ material into his manuscript, he added hand-
CrossCheck, allow these sections of the paper written notations detailing where the text
to be excluded from the search, together with came from. These notations were supposed to
any text enclosed in quotation marks. have been printed in the body of his chapter’.
If systematic reviews or databases are Nevertheless he was found guilty of ‘grossly
updated, the original authors may be replaced negligent scholarship’ and resigned as chair-
by others. An updated review or database man of the Department of Medicine.4


Defining and responding to plagiarism 39

Authors who admit intentional copying recommended by the COPE flowcharts based
may nevertheless insist that this is acceptable on experience that apparently ‘cut-and-dried’
in their discipline or culture and that, rather cases of misconduct often turn out to be more
than representing academic theft or laziness, complex than they appear on the surface.
it is, in fact, a form of flattery or ‘homage’ If an editor detects copying in a manuscript
to the original author. They may also sug- that is going to be rejected, then contacting
gest that quotation marks are unnecessary a head of department or dean might prevent
because specialist readers (for whom they are the authors from simply submitting the manu-
writing) will immediately recognize the quota- script, unchanged, to another journal.
tions and be aware of their source. However,
editors and readers may find these arguments Language
Text-matching software will only detect text
copying in the same language. However,
Author seniority
republication of an unattributed translation
Since editors may believe that some forms of another person’s work is also plagiarism,
of plagiarism result from poor mentorship or although it is harder to detect and may be
supervision rather than intentional miscon- harder to prove unless extensive.
duct, their response may vary according to Just as editors’ responses may depend on republication of
the seniority of the authors involved. Editors the authors’ seniority, they may also depend
may apply different sanctions to junior authors on whether authors are writing in their native
an unattributed
who they believe genuinely did not realize language since editors recognize the difficulties translation
they were doing something inappropriate from that non-native writers face in correctly para- of another
those applied to experienced researchers who phrasing other authors’ work. In some cases, person’s work is
are expected to know better. Thus, an editor researchers may actually have been encour-
may respond to the copying of a paragraph aged, when learning a language, to adapt sen- also plagiarism,
from a cited paper by asking a junior author tences and ‘borrow’ structures from published although it is
to paraphrase (if detected before publication) works. This may result in so-called ‘patch’ (or harder to detect
or issue a correction (if detected after publica- ‘patchwork’) writing.5 This form of copying
tion). However, for a similar degree of copy- will only be detected by sensitive text-match-
ing by a senior author, the same editor might ing systems and those that employ a degree
reject or retract a submission and consider of ‘fuzzy’ matching, since authors are likely to
informing the author’s institution. have changed some words in adapting the sen-
Informing an author’s institution is gener- tence for their own use. Authors who use this
ally considered to be a relatively serious action technique usually copy from a wide range of
to take, since it may have grave consequences sources, often with individual sentences com-
for the researcher concerned. Editors there- ing from different publications. This may result
fore tend to be reluctant to inform institu- in a high total similarity ‘score’ for the article
tions except in serious cases of misconduct from an anti-plagiarism detection system such
and when they feel they have well-founded as CrossCheck, but the matched text will be
suspicions of wrong-doing. However, if con- found to come from many sources, and each
tacting an institution is viewed, not as a copied section will be short (with few or no
potential punishment for the author, but as an substantial chunks of copied text). However,
attempt to engage the institution in dialogue few, if any, of the sources of the copied text are
and work together to prevent future prob- likely to be cited in the publication, since they
lems, one might argue that editors should con- may be on unrelated topics.
tact institutions more often and definitely in Some editors may see little harm in authors
cases where they feel junior researchers have who describe their own methods and find-
received inadequate training or guidance, ings accurately, but use sentence structures
since this is something the institution may be taken from other publications. Others may
able to remedy. Before contacting an institu- regard this as a sign of poor scholarship or a
tion it is advisable to ask the author(s) for an form of minor plagiarism. The acceptability
explanation even in the face of strong evi- of ‘patch’ writing probably depends on the
dence of apparently blatant plagiarism. This is originality of the writing being copied. While


40 Defining and responding to plagiarism

it may be entirely unacceptable for works of Possible responses to misconduct

creative fiction, it may be considered accept-
able when describing widely used methods (From COPE, Guidelines on Good Publication
which, as already mentioned, may require a Practice, 1999)
degree of standardized text to ensure accuracy. The following [sanctions] are ranked in
If the copied structures are clear and gram- approximate order of severity:
matically correct, some editors may even feel (1) A letter of explanation (and education)
that this method of writing will benefit readers to the authors, where there appears
and journals, since methods will be accurately to be a genuine misunderstanding of
described and the manuscript will require less principles.
copyediting to correct grammatical mistakes. (2) A letter of reprimand and warning as to
However, others may have concerns that future conduct.
authors will be tempted to copy inappropri- (3) A formal letter to the relevant head of
ate phrases that do not correctly describe their institution or funding body.
own research, especially if they do not com- (4) Publication of a notice of redundant
pletely understand the phrases being copied. publication or plagiarism.
To misrepresent research methods (e.g. by (5) An editorial giving full details of the
stating that a study was prospective or ran- misconduct.
domized when, in fact, it was not) is gener- (6) Refusal to accept future submissions
ally considered a serious form of misconduct. from the individual, unit, or institution
Editors may therefore be concerned that, if responsible for the misconduct, for a
they tolerate ‘patch’ writing, such misrepresen- stated period.
screening incurs tation may go undetected. (7) Formal withdrawal or retraction of the
costs, and paper from the scientific literature,
therefore editors Detecting and responding to plagiarism informing other editors and the indexing
and publishers The original COPE guidelines on good publi- (8) Reporting the case to the General
need to decide cation practice (published in 1999) noted that Medical Council, or other such author-
the best ways of ‘plagiarism ranges from the unreferenced use ity or organisation which can investigate
of others’ published ideas … to submission and act with due process.
employing it under “new” authorship of a complete paper,
sometimes in a different language’.6 However,
decide the best ways of employing it. The
these guidelines did not describe what editors
options include:
should do if they encountered these different
forms. The guidelines did offer general guid- • screening all manuscripts on receipt
ance on the sanctions that editors might take • screening manuscripts that are sent out for
against authors (see text box). external peer review
The COPE flowcharts on plagiarism (pub- • screening manuscripts that are provision-
lished in 2006) recommend different responses ally accepted
for ‘clear plagiarism’ and ‘minor copying’ • screening a random sample of manuscripts
but provide only rather general indications • using the software only in cases when pla-
of how editors might distinguish these two giarism is suspected.
Anecdotal reports suggest that, when
CrossCheck was first available, many journals
Screening for plagiarism started by screening only accepted manuscripts
The availability of powerful tools such as but later switched to screening all submissions
CrossCheck makes it possible to screen sub- because of the frequency of problems they dis-
missions for matching text and some journals covered (since they did not wish to send peer-
are now doing this routinely. However, screen- reviewers manuscripts that were later found
ing incurs costs (in the form of charges for to raise concerns about plagiarism or redun-
using the tools, and in terms of editorial time), dancy). The Editor of Anesthesia & Analgesia
and therefore editors and publishers need to noted in an editorial that ‘I have screened


Defining and responding to plagiarism 41

every submitted manuscript for many months. Considering the COPE flowcharts, maybe they
Approximately 1 of every 10 submissions has should provide more guidance to help editors
had unacceptable amounts of text taken ver- distinguish major from minor plagiarism. One
batim and without attribution from another possibility would be to produce definitions
source.’ Elsevier has announced that, as from based on the characteristics described above.
2014, it will use CrossCheck on all submis- For example, major plagiarism could be
sions and has integrated the text-matching defined as any case involving:
software into its editorial system.7,8 • unattributed copying of another person’s
data/findings, or
Defining plagiarism • resubmission of an entire publication under
another author’s name (either in the origi- clear-cut cases
Editors need to decide how to interpret and nal language or in translation), or of serious
respond to findings of text similarity. It is • verbatim copying of >100 words of original plagiarism
important that authors receive fair and con- material in the absence of any citation to
sistent treatment from journals, but devising the source material, or
(e.g. whole
a detailed policy on responses to plagiarism is • unattributed use of original, published aca- articles or large
difficult given the many forms that plagiarism demic work, such as the structure, argu- sections of text)
can take. Because text-matching software has ment, or hypothesis/idea of another person
only become available relatively recently, cases
may warrant
or group where this is a major part of the
of plagiarism are very likely to be uncovered new publication and there is evidence that retractions
in back issues of the journal. Editors therefore it was not developed independently.
need a clear policy for responding to plagia-
rism in material published recently and further Minor plagiarism could be defined as:
back in the past. • verbatim copying of <100 words without
Clear-cut cases of serious plagiarism (e.g. indicating that these are a direct quotation
whole articles or large sections of text) may from an original work (whether or not the
warrant retractions. Since the general concept source is cited), unless the text is accepted
of plagiarism is not new, and large-scale pla- as widely used or standardized (e.g. the
giarism has been identified as a serious form description of a standard technique)
of misconduct for decades, most editors would • close copying (not quite verbatim, but
agree that this is the correct course of action. changed only slightly from the original) of
However, identification of ‘patch writing’ or significant sections (e.g. >100 words) from
‘micro-plagiarism’ has only become possible another work (whether or not that work is
with the availability of specialized software. cited).
Some editors may therefore feel uncomfort- Use of images without acknowledgement of
able about applying sanctions to authors retro- the source could be defined as:
spectively. One solution to this problem would
be to announce an amnesty for older publica- • republication of an image (photograph, dia-
tions (i.e. an agreement that the journal will gram, drawing, etc.) generated by another
person without acknowledging the source.
not take action if minor plagiarism is found
in previous issues) but warning authors that Journal responses could then be matched to
text similarity in future submissions will not be these. For example:
To devise workable policies for both sub- • Minor plagiarism in a submitted article – write
mitted and published articles, editors need to to author and request reworking or (if arti-
cle is being rejected) point out that minor
consider the thresholds for deciding when to:
plagiarism has been detected and advising
• educate authors and ask them to rewrite the authors that this should be corrected
copied text before resubmission.
• reject an article • Minor plagiarism in a published article – con-
• issue a correction (for a published article) tact author and discuss findings, and issue
• issue a retraction (for a published article) a correction and apology. It is important to
• inform an author’s institution. ensure that corrections are clearly indicated


42 Defining and responding to plagiarism

in all versions of the publication (i.e. both tools such as CrossCheck but publishers need
online and print) and, where possible, via to decide how best to use these. Two studies
databases and indexing systems – the use of presented at the 2013 Peer Review Congress
systems such as CrossMark offers possibili- assessed the amount of time needed to screen
ties for alerting readers to corrections even articles.9,10 Both concluded that screening
if they are using a downloaded PDF. was feasible and useful, but also noted that
• Major plagiarism in a submitted article – pres- it required an investment of time which, for
ent findings to all authors and ask them journals receiving large numbers of submis-
policies need to respond; ask the authors if all or only sions, could be substantial.
some of them are responsible for the pla-
a workable giarized sections, decide if any authors References
definition of were unaware of the plagiarism and, if so,
1. Bailey, J. The world’s first plagiarism case. http://www.pla-
plagiarism. This whether they are in any way responsible for
the behaviour of the other authors (e.g. in
article offers a supervisory capacity); explain that plagia-
first-plagiarism-case/ (accessed 2 October 2013).
2. Committee on Publication Ethics. Flowcharts
some possible rism is unacceptable and that you plan to ‘What to do if you suspect plagiarism’
suggestions inform their institution; contact the institu-
tions of authors you consider were directly Submitted.pdf and
u2/02B_Plagiarism_Published.pdf. 2008 (accessed 2
involved with, or should take responsibility October 2013).
for, the plagiarism 3. Committee on Publication Ethics. Retraction guidelines.
• Major plagiarism in a published article – as for
submitted article, then retract article. lines.pdf (accessed 2 October 2013).
• Use of images without acknowledgement of 4. Norman, C. 1984. Stanford investigates plagiarism
the source – if the image contains data from charge. Science, 224: 35–36.
5. Kerans, M.E. and de Jager, M. 2010. Handling plagia-
another person’s research (e.g. a graph), rism at the manuscript editor’s desk. European Science
and this is shown as if it were the work of Editing, 36: 62–66.
the copyist, this should be treated as data 6. Committee on Publication Ethics. Guidelines on Good
copying (i.e. major plagiarism). For images Publication Practice. 1999.
that do not contain original data (e.g. dia- annualreport/1999
grams showing processes, maps, illustrative 7. Kleikamp, E. CrossCheck–EES integration go-
live date announced. 10 September 2013.
photographs), the author of a submitted
paper should be told to seek permission for cations/crosscheck- ees-integration-go -live - date -
republication from the copyright holder, announced/ (accessed 2 October 2013).
remove images for which permission is not 8. Wager, E. High-tech approaches to high-tech fraud.
granted, and insert appropriate acknowl- Elsevier Connect, 15 May 2013.
edgements for images for which permis- nect/high-tech-approaches-to-high-tech-fraud (accessed
2 October 2013).
sion has been granted; if such images have 9. Vermette, H.L., Benner, R.S., and Scott, J.R. Value of
already been republished, the editor should plagiarism detection for manuscripts submitted to a
contact the author and issue a correction medical specialty journal. http://www.peerreviewcon-
giving the appropriate acknowledgements. (accessed 2 October 2013).
10. Flavall, E., Barbour, V., Bernstein, R., Hickling, K., and
Morris, M. Implementation of plagiarism screening for
Conclusions PLOS journals.
index.html (accessed 2 October 2013).
Editors and publishers should consider their
policies on detecting and handling plagiarism. Elizabeth Wager
To be helpful, and to be applied consistently, Publications Consultant
such policies need a workable definition of Sideview
plagiarism. This article offers some possible 19 Station Road
suggestions. Text-matching software coupled Princes Risborough HP27 9DE, UK
with extensive databases can create powerful Email: