Salami-slicing and plagiarism: How should we respond? · Salami-licing and plagiaim: Ho hold e...
Transcript of Salami-slicing and plagiarism: How should we respond? · Salami-licing and plagiaim: Ho hold e...
Vol.:(0123456789)
Advances in Health Sciences Education (2019) 24:3–14https://doi.org/10.1007/s10459-019-09876-7
1 3
EDITORIAL
Salami‑slicing and plagiarism: How should we respond?
Martin G. Tolsgaard1 · Rachel Ellaway2 · Nikki Woods3 · Geoff Norman4
© Springer Nature B.V. 2019
One particularly odious task that every journal editor must confront from time to time is the investigation of a case of scientific fraud. We are not talking about blatant cases, where data are falsified; these are almost impossible to detect. Rather, a far more common occur-rence is the deliberate use of previously published date and/or text—either the author’s own or, more seriously, plagiarized from other authors’ publications. Three labels are attached to these acts: Plagiarism—copying from another author’s published work; Auto-plagiarism—duplication of previously published by the authors without clearly stating that this was the case; Salami-slicing—authors have published parts of a study in multiple papers instead of providing the full story in a single paper. These are not new problems in medical education research (Brice et al. 2009) but the incidence seems to be increasing over the years (Steen 2011). Although these problems have been addressed in recent edito-rials in medical education journals (Norman 2014; Eva 2017), there seems to be a lack of consensus on how to handle manuscripts that have these problems.
Identifying examples of text copied from elsewhere used to be difficult, as it required extensive cross-referencing. Nowadays, plagiarism detection tools feature in many jour-nals’ workflows (including AHSE), so previously published text can be quickly identi-fied along with its provenance. However, while these automated tools can indicate where sentences or fragments arise from other texts, many of these are perfectly legitimate. It is almost impossible to have a paper with no “plagiarized” text based on automated search, and judgment is required to identify the demarcation between legitimate uses of phrases or sentences and clear plagiarism. Similar issues arise with “auto-plagiarism” and salami-slicing. When an author uses similar methodology in a series of studies, it is natural that some methodological description may well be common to several papers, Salami slicing for one editor might be seen as programmatic research by others (Eva 2017). Nevertheless, editors need to determine if there is clear evidence of misconduct, whether there are rea-sons to sanction the authors for this conduct, and if so then how this should be done.
* Geoff Norman [email protected]
1 Copenhagen Academy for Medical Education and Simulation (CAMES), Rigshospitalet and University of Copenhagen, Copenhagen, Denmark
2 Department of Community Health Sciences, University of Calgary, Calgary, AB, Canada3 Department of Family and Community Medicine, University of Toronto, Toronto, ON, Canada4 Department of Health Research Methods, Evidence and Impact, McMaster University, Hamilton,
ON, Canada
4 M. G. Tolsgaard et al.
1 3
About 2 years ago, we began to systematically track suspected incidents. To date, 25 manuscripts have been submitted to AHSE where editors have suspected plagiarism (15), auto-plagiarism (5) or salami-slicing (5). Thirteen authors received a warning, 4 manu-scripts have been rejected cautioning authors and 8 authors have been blacklisted and not allowed to submit any more manuscripts to the Journal. Of course, it remains to be seen how many of these articles would not be accepted even without issues of plagiarism. How-ever, all of these decisions were made by the editor-in-chief without consultation, and it remains to be seen what kind of agreement would arise among editors. Guidelines do exist, (COPE) (https ://publi catio nethi cs.org/resou rces/flowc harts ) and articles have appeared in other medical education journals (Brice et al. 2009; Eva 2017; Norman 2014).
To inform journal policy and to provide author guidance for Advances in Health Sci-ences Education (AHSE), we surveyed our editorial board with respect to how members of the editorial board considered salami-slicing, plagiarism, and auto-plagiarism in terms of scientific conduct and its appropriate consequences.
A study
We designed an online-survey of possible practices related to these areas and potential actions. Survey items were developed and informed by existing literature on salami-slicing, auto-plagiarism, and plagiarism in health professions education and in accordance with the guidelines for survey development described by Gehlbach et al. (2010). Survey items were generated based on consensus among the study authors and were categorized in three parts; seriousness of action, expected response from editors, and transparency declaration. The survey was sent to all associate editors of Advances in Health Sciences Education. Survey results were presented at the editorial board meeting in August 2018. From this we devel-oped a second questionnaire that included items that involved editor responsibilities and the role of pre-publication of scientific manuscripts. This second questionnaire was sent to AHSE editors in September 2018. Editors received one reminder per questionnaire within 4 weeks of the invitation. Descriptive statistics were performed and extent of plagiarism, auto-plagiarism, salami-slicing as well as their consequences were compared using para-metric statistics, when appropriate. Bonferroni corrections were used to adjust for multiple comparisons.
Ethical approval was granted at the University of Calgary (REB17-2407). Survey stud-ies are exempt from ethical approval in Denmark according to national regulations.
Twenty-four of the 25 AHSE associate editors completed both survey rounds (96%). Substantial plagiarism and auto-plagiarism were perceived as ‘a major issues’ or ‘unaccep-table academic behavior’ by at least half of the editors when authors failed to provide any referencing to the publication, from which the text was plagiarized (Table 1). On the con-trary, if authors provided appropriate referencing, the majority of editors perceived varying degrees of plagiarism and auto-plagiarism to be ‘very minor’ or ‘minor issues’ (Table 1).
Acceptable behaviours
Plagiarism was less tolerated than auto-plagiarism when authors failed to provide adequate referencing (t = −4.3, d.f. = 19; p = 0.002 for ‘substantial parts of the text’ and t = −5.0, d.f. = 20; p < 0.01 for ‘sentences’). However, there were no differences in editors’ accept-ance of plagiarism versus auto-plagiarism when authors provided appropriate referencing
5Salami-slicing and plagiarism: How should we respond?
1 3
Tabl
e 1
Dist
ribut
ion
of e
dito
rs’ r
atin
g of
serio
usne
ss
‘Ref
eren
cing
’ inc
lude
s (1
) flag
ging
whi
ch p
art o
f the
text
has
bee
n pr
evio
usly
pub
lishe
d (e
.g. u
sing
quo
tatio
n m
arks
) and
(2) c
iting
the
sour
ce fr
om w
hich
the
text
has
bee
n co
pied
This
is a
ver
y m
inor
issu
e (%
)
Min
or is
sue—
note
in
adj
udic
atio
n (%
)Th
is is
a
maj
or is
sue
(%)
Una
ccep
tabl
e ac
adem
ic b
ehav
iour
(%
)
Rega
rdin
g au
to-p
lagi
aris
m (c
opyi
ng fr
om o
wn
work
)Su
bsta
ntia
l par
ts o
f the
of t
ext h
ave
appe
ared
in th
e au
thor
s’ p
rior p
ublic
atio
n W
ITH
OU
T an
y re
fere
ncin
g4.
760.
0061
.90
33.3
3
Subs
tant
ial p
arts
of t
he o
f tex
t hav
e ap
pear
ed in
the
auth
ors’
prio
r pub
licat
ion
WIT
H a
ppro
pria
te
refe
renc
ing
33.3
342
.86
19.0
54.
76
Sent
ence
s or p
hras
es o
f tex
t hav
e ap
pear
ed in
the
auth
ors’
prio
r pub
licat
ion
WIT
H a
ppro
pria
te
refe
renc
ing
63.6
422
.73
9.09
4.55
Rega
rdin
g pl
agia
rism
(cop
ying
from
oth
ers’
wor
k)Su
bsta
ntia
l par
ts o
f the
of t
ext h
ave
been
take
n fro
m p
rior p
eer r
evie
wed
pub
licat
ion
WIT
HO
UT
any
refe
renc
ing
0.00
0.00
5.00
95.0
0
Subs
tant
ial p
arts
of t
he o
f tex
t hav
e be
en ta
ken
from
prio
r pee
r rev
iew
ed p
ublic
atio
n W
ITH
app
ro-
pria
te re
fere
ncin
g25
.00
30.0
045
.00
0.00
Sent
ence
s of t
ext h
ave
been
take
n fro
m p
rior p
eer r
evie
wed
pub
licat
ion
WIT
HO
UT
any
refe
renc
ing
0.00
19.0
523
.81
57.1
4Re
gard
ing
sala
mi-s
licin
g (p
rese
ntin
g pr
evio
usly
pub
lishe
d m
ater
ial)
The
over
all i
dea
or re
sear
ch q
uesti
on is
sim
ilar t
o pu
blic
atio
ns b
y th
e sa
me
auth
or o
r gro
up o
f au
thor
s but
the
data
set i
s new
. The
aut
hors
HAV
E re
fere
nced
thei
r pre
viou
s wor
k61
.90
28.5
79.
520.
00
The
over
all i
dea
or re
sear
ch q
uesti
on is
sim
ilar t
o pr
evio
us p
ublic
atio
ns b
y th
e sa
me
auth
or o
r gro
up
of a
utho
rs b
ut th
e da
tase
t is n
ew. T
he a
utho
rs h
ave
NO
T re
fere
nced
thei
r pre
viou
s wor
k4.
7628
.57
42.8
623
.81
The
data
-set
is th
e sa
me
as u
sed
in a
pre
viou
s stu
dy b
ut th
e re
sear
ch q
uesti
on is
new
. The
aut
hors
H
AVE
refe
renc
ed th
eir p
revi
ous w
ork
71.4
328
.57
0.00
0.00
6 M. G. Tolsgaard et al.
1 3
(t = −1.2, d.f. = 19; p = 1.00 for ‘substantial parts of the text’ and t = 2.7, d.f. = 20; p = 1.00 for ‘sentences. Similarly, editors perceived salami-slicing to be a serious matter when authors failed to reference previous publications, but much less so if the prior work was appropriately referenced and difference with present work clearly described.
All editors agreed that publishing abstracts in online conference proceedings was either a minor issue or completely acceptable, and 90% (22) agreed that reusing material from theses was also acceptable. However, there was more variation in editors’ judgment regard-ing ‘pre-publicizing’, with the majority of editors being of the opinion that pre-publication is a ‘minor problem’ or not a problem at all. Free-text responses revealed concerns regard-ing pre-publication for several reasons, including ethical considerations (breaking the rules of publication by submitting several places), legal considerations (concerns regarding who owns the manuscript), business considerations (pre-publicizing threatens the busi-ness model of the journal), and quality concerns (in particular the absence of peer review for pre-published material, which makes it difficult for consumers to tell the difference between content published in real journals versus pre-published papers).
Consequences
A small proportion of editors (0–15%) recommended that there should be serious con-sequences (noted in Table 2) in cases of auto-plagiarism or salami-slicing regardless of whether or not authors provided adequate referencing. Just under half of the respondents recommended serious consequences for plagiarism if authors failed to provide adequate referencing, whereas only one respondent recommended retraction (if already published) in the case of adequate referencing.
Declarations
86% of the respondents agreed that authors should declare potential problems regard-ing salami-slicing, plagiarism, and auto-plagiarism in the cover letter to editors and 91% thought it should be mentioned in the manuscript text. None of the editors indicated that omitting such declarations was acceptable.
Editorial responsibility
43% of editors believed that either general or specific rules are needed to guide editors in questions regarding plagiarism and salami-slicing with latitude for individual assessment. 91% felt they had a shared responsibility with the editorial team to assess whether or not authors have plagiarized or salami-sliced their paper, the other two did not see this as their job and felt responsible only for the academic content.
Discussion
AHSE associate editors considered auto-plagiarism, plagiarism, and salami-slicing to be a problem in academic publishing. Of the three, plagiarism was considered a more serious issue. The severity of the problem depended on the magnitude of overlap and the extent to which they had or had not declared that they were doing so. There was
7Salami-slicing and plagiarism: How should we respond?
1 3
Tabl
e 2
Edi
tors
’ rat
ings
of a
ppro
pria
te c
onse
quen
ces.
Serio
us c
onse
quen
ces
incl
ude:
Bla
cklis
t for
futu
re s
ubm
issi
ons,
Not
ify p
ublic
atio
n w
here
mat
eria
l orig
inal
ly p
ublis
hed,
N
otify
all
mai
nstre
am jo
urna
ls, N
otify
aut
hors
’ dea
ns, d
epar
tmen
t cha
irs, A
sk fo
r ret
ract
ion
(if a
lread
y pu
blis
hed)
No
addi
tiona
l re
spon
se (%
)Re
quire
aut
hor
to ju
stify
thei
r ac
tions
(%)
Cau
tion
auth
or
(%)
Reje
ct p
aper
w
ith e
xpla
na-
tion
(%)
Bla
cklis
t for
fu
ture
subm
is-
sion
s (%
)
Not
ify p
ublic
a-tio
n w
here
m
ater
ial o
rigi-
nally
pub
lishe
d (%
)
Not
ify a
ll m
ain-
strea
m jo
urna
ls
(%)
Not
ify a
utho
rs’
dean
s, de
part-
men
t cha
irs (%
)
Ask
for r
etra
c-tio
n (if
alre
ady
publ
ishe
d) (%
)
Rega
rdin
g au
to-p
lagi
aris
m (s
elec
t all
that
app
ly)
Subs
tant
ial p
arts
of
the
of te
xt
have
bee
n ta
ken
from
the
auth
ors’
prio
r pu
blic
atio
n W
ITH
OU
T an
y re
fere
nc-
ing
5.00
30.0
050
.00
75.0
015
.00
20.0
05.
0010
.00
15.0
0
Subs
tant
ial p
arts
of
the
of te
xt
have
bee
n ta
ken
from
in
the
auth
ors’
pr
ior p
ublic
a-tio
n W
ITH
ap
prop
riate
re
fere
ncin
g
30.0
055
.00
45.0
010
.00
0.00
0.00
0.00
0.00
5.00
Sent
ence
s of
text
hav
e be
en
take
n fro
m
the
auth
ors’
pr
ior p
ublic
a-tio
n W
ITH
ap
prop
riate
re
fere
ncin
g
76.1
923
.81
14.2
94.
760.
000.
000.
000.
000.
00
8 M. G. Tolsgaard et al.
1 3
Tabl
e 2
(con
tinue
d) No
addi
tiona
l re
spon
se (%
)Re
quire
aut
hor
to ju
stify
thei
r ac
tions
(%)
Cau
tion
auth
or
(%)
Reje
ct p
aper
w
ith e
xpla
na-
tion
(%)
Bla
cklis
t for
fu
ture
subm
is-
sion
s (%
)
Not
ify p
ublic
a-tio
n w
here
m
ater
ial o
rigi-
nally
pub
lishe
d (%
)
Not
ify a
ll m
ain-
strea
m jo
urna
ls
(%)
Not
ify a
utho
rs’
dean
s, de
part-
men
t cha
irs (%
)
Ask
for r
etra
c-tio
n (if
alre
ady
publ
ishe
d) (%
)
Sent
ence
s of
text
hav
e be
en
take
n fro
m th
e au
thor
s’ p
rior
publ
icat
ion
WIT
HO
UT
any
refe
renc
-in
g
4.76
38.1
071
.43
23.8
10.
004.
760.
004.
764.
76
Rega
rdin
g pl
agia
rism
(sel
ect a
ll th
at a
pply
)Su
bsta
ntia
l par
ts
of th
e of
text
ha
s app
eare
d in
prio
r pee
r re
view
ed
publ
icat
ion
WIT
HO
UT
any
refe
renc
-in
g
0.00
10.0
015
.00
90.0
065
.00
45.0
035
.00
40.0
040
.00
Subs
tant
ial p
arts
of
the
of te
xt
has a
ppea
red
in p
rior p
eer
revi
ewed
pub
-lic
atio
n W
ITH
ap
prop
riate
re
fere
ncin
g
0.00
45.0
060
.00
30.0
00.
000.
000.
000.
000.
00
9Salami-slicing and plagiarism: How should we respond?
1 3
Tabl
e 2
(con
tinue
d) No
addi
tiona
l re
spon
se (%
)Re
quire
aut
hor
to ju
stify
thei
r ac
tions
(%)
Cau
tion
auth
or
(%)
Reje
ct p
aper
w
ith e
xpla
na-
tion
(%)
Bla
cklis
t for
fu
ture
subm
is-
sion
s (%
)
Not
ify p
ublic
a-tio
n w
here
m
ater
ial o
rigi-
nally
pub
lishe
d (%
)
Not
ify a
ll m
ain-
strea
m jo
urna
ls
(%)
Not
ify a
utho
rs’
dean
s, de
part-
men
t cha
irs (%
)
Ask
for r
etra
c-tio
n (if
alre
ady
publ
ishe
d) (%
)
Sent
ence
s of
text
has
ap
pear
ed in
pr
ior p
eer
revi
ewed
pu
blic
atio
n W
ITH
OU
T an
y re
fere
nc-
ing
0.00
38.1
047
.62
66.6
719
.05
19.0
514
.29
19.0
523
.81
Sent
ence
s of
text
has
ap
pear
ed in
pr
ior p
eer
revi
ewed
pub
-lic
atio
n W
ITH
ap
prop
riate
re
fere
ncin
g
52.3
833
.33
28.5
79.
520.
000.
000.
000.
004.
76
10 M. G. Tolsgaard et al.
1 3
Tabl
e 2
(con
tinue
d) No
addi
tiona
l re
spon
se (%
)Re
quire
aut
hor
to ju
stify
thei
r ac
tions
(%)
Cau
tion
auth
or
(%)
Reje
ct p
aper
w
ith e
xpla
na-
tion
(%)
Bla
cklis
t for
fu
ture
subm
is-
sion
s (%
)
Not
ify p
ublic
a-tio
n w
here
m
ater
ial o
rigi-
nally
pub
lishe
d (%
)
Not
ify a
ll m
ain-
strea
m jo
urna
ls
(%)
Not
ify a
utho
rs’
dean
s, de
part-
men
t cha
irs (%
)
Ask
for r
etra
c-tio
n (if
alre
ady
publ
ishe
d) (%
)
Rega
rdin
g sa
lam
i-slic
ing
(sel
ect a
ll th
at a
pply
)Th
e ov
eral
l ide
a or
rese
arch
qu
estio
n is
ba
sica
lly
the
sam
e as
in
pre
viou
s pu
blic
a-tio
ns b
y th
e sa
me
auth
or
or g
roup
of
auth
ors b
ut
the
data
set
is n
ew. T
he
auth
ors H
AVE
refe
renc
ed
thei
r pre
viou
s w
ork
33.3
366
.67
14.2
90.
000.
000.
000.
000.
000.
00
11Salami-slicing and plagiarism: How should we respond?
1 3
Tabl
e 2
(con
tinue
d) No
addi
tiona
l re
spon
se (%
)Re
quire
aut
hor
to ju
stify
thei
r ac
tions
(%)
Cau
tion
auth
or
(%)
Reje
ct p
aper
w
ith e
xpla
na-
tion
(%)
Bla
cklis
t for
fu
ture
subm
is-
sion
s (%
)
Not
ify p
ublic
a-tio
n w
here
m
ater
ial o
rigi-
nally
pub
lishe
d (%
)
Not
ify a
ll m
ain-
strea
m jo
urna
ls
(%)
Not
ify a
utho
rs’
dean
s, de
part-
men
t cha
irs (%
)
Ask
for r
etra
c-tio
n (if
alre
ady
publ
ishe
d) (%
)
The
over
all i
dea
or re
sear
ch
ques
tion
is
basi
cally
th
e sa
me
as
in p
revi
ous
publ
ica-
tions
by
the
sam
e au
thor
or
gro
up o
f au
thor
s but
the
data
set i
s new
. Th
e au
thor
s ha
ve N
OT
refe
renc
ed
thei
r pre
viou
s w
ork
0.00
47.6
252
.38
52.3
89.
524.
764.
764.
7614
.29
12 M. G. Tolsgaard et al.
1 3
Tabl
e 2
(con
tinue
d) No
addi
tiona
l re
spon
se (%
)Re
quire
aut
hor
to ju
stify
thei
r ac
tions
(%)
Cau
tion
auth
or
(%)
Reje
ct p
aper
w
ith e
xpla
na-
tion
(%)
Bla
cklis
t for
fu
ture
subm
is-
sion
s (%
)
Not
ify p
ublic
a-tio
n w
here
m
ater
ial o
rigi-
nally
pub
lishe
d (%
)
Not
ify a
ll m
ain-
strea
m jo
urna
ls
(%)
Not
ify a
utho
rs’
dean
s, de
part-
men
t cha
irs (%
)
Ask
for r
etra
c-tio
n (if
alre
ady
publ
ishe
d) (%
)
The
data
-set
is
the
sam
e as
use
d in
a
prev
ious
stu
dy b
ut
the
rese
arch
qu
estio
n is
ne
w. T
he
auth
ors H
AVE
refe
renc
ed
thei
r pre
viou
s w
ork
52.3
852
.38
9.52
4.76
0.00
0.00
0.00
0.00
0.00
The
data
-set
is
the
sam
e as
use
d in
a
prev
ious
stu
dy b
ut th
e re
sear
ch q
ues-
tion
is n
ew.
The
auth
ors
have
NO
T re
fere
nced
th
eir p
revi
ous
wor
k
0.00
52.3
857
.14
57.1
49.
524.
764.
764.
769.
52
13Salami-slicing and plagiarism: How should we respond?
1 3
less agreement regarding the severity of consequences that should follow different types of scientific misconduct as well as what should be considered scientific misconduct to begin with. For instance, some editors considered pre-publishing a paper as equivalent to plagiarism whereas other editors did not see it as a problem at all.
The majority of AHSE editors were of the opinion that it is a shared responsibility of the editorial team to carefully evaluate manuscripts for potential scientific miscon-duct. They expressed a wish to operate within some rules with latitude for individual assessment based on context. However, current guidelines for the assessment and practi-cal handling of scientific misconduct (COPE) may not provide the guidance needed by editors in our field. Instead, some principles may be deducted from our survey results, which may help guide editors as well as authors:
• Authors must declare potential problems with plagiarism, auto-plagiarism and salami-slicing in the cover letter to the editors as well as in the manuscript text.
• Serious consequences (such as blacklisting, contacting authors’ deans/department chairs, or retracting published manuscripts) will be considered when authors plagia-rize and fail to be transparent about it.
• Auto-plagiarism and salami-slicing on the other hand may result in rejection and cautioning of the authors but should not automatically result in serious consequences such as those mentioned above.
• Reuse of authors’ own conference abstracts or text from previously published theses is considered acceptable practice.
• The consequence of plagiarism, auto-plagiarism and salami-slicing should be evalu-ated individually based on the amount of text or material involved, the extent to which this is acknowledged, and on the type of misconduct committed (with plagia-rism being considered more serious than auto-plagiarism and salami-slicing).
• Collectively, editorial teams are responsible for detection of these issues and for for-mulating an appropriate response.
Given the somewhat fluid and evolving nature of academic publishing, these principles are intended to guide practice and to some extent normalize it, at least within this jour-nal, but they are not expected to replace editorial judgement. We advance them as a guide to those wishing to publish in this journal and as the basis for ongoing debate on what constitutes scientific misconduct in health professions education.
References
Brice, J., Bligh, J., Bordage, G., Colliver, J., Cook, D., Eva, K. W., et al. (2009). Publishing ethics in medical education journals. Academic Medicine, 84(10 Suppl), S132-4.
COPE Guidelines. Retrieved January 15, 2019 from https ://publi catio nethi cs.org/resou rces/flowc harts .Eva, K. W. (2017). How would you like your salami? A guide to slicing. Medical Education, 51(5),
456–457.Gehlbach, H., Artino, A. R., Jr., & Durning, S. (2010). AM last page: Survey development guid-
ance for medical education researchers. Academic Medicine, 85(5), 925. https ://doi.org/10.1097/ACM.0b013 e3181 dd3e8 8.
Norman, G. (2014). Data dredging, salami-slicing, and other successful strategies to ensure rejection: Twelve tips on how to not get your paper published. Advances in Health Sciences Education Prac-tice, 19(1), 1–5.
14 M. G. Tolsgaard et al.
1 3
Steen, R. G. (2011). Retractions in the scientific literature: Is the incidence of research fraud increasing? Journal of Medical Ethics, 37(4), 249–253.
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.