Successful Letters in Physical Review Letters: An editor's perspective Manolis Antonoyiannakis

Successful Letters
in Physical Review Letters:
An editor's perspective
Manolis Antonoyiannakis
Physical Review Letters
[email protected]
24 September 2008
Physics Department, University of Tokyo
Greetings from All at PRL
The APS editors are international
From 24 countries
…but we have still not had the good fortune
to have a Japanese colleague join us!
a dynamic and personable colleague
who has a Ph.D. and some postdoctoral experience
particle, nuclear, astrophysics, or quantum information
www.aps.org.
1958 - 2008
Physical Review Letters’
First 50 Years:
Birth, Growth and Success
Letters to
the Editor
1929
Papers Published
In The Physical Review
APS
take over
1913
Tate
1926
By mid 30’s, PR became the journal in
Physics, even for European physicists
Sam A. Goudsmit
(1902 - 1978)
With the APS journals:1951 - 1974
Physical Review Letters
Sam Goudsmit’s Experiment
1958
Brookhaven National Lab
Sam A. Goudsmit Rapid publication of very important results
(1902 - 1978)
• speed: 6 weeks reduced to 3 weeks
• Little or no refereeing
• An experiment (success or failure?)
Success
Moving Physics Forward
Milestone Papers: http://prl.aps.org/50years/milestones
• Papers that report major discoveries
• Papers that move an entire field
• Papers that create a whole new research area
• Papers that lead to cross-field fertilization
Nobel Prizes
Nobel Prize Winning Papers in PRL*
Physics
Decade
1976
Chemistry
1970’s
1973
1979
1980’s
1980 (1982) 1985 1988 (1989)
1990’s
1990 (1993, 1994) 1995 1997 1998
1998
2000’s
(2001) 2002 2004 2005 (2006) 2007
2000
* Counting may not be complete
PRL 50TH ANNIVERSARY
New On-line Content: http://prl.aps.org/50years
• Timeline
• Milestone Letter each week:
• Editorials and Essays:
Which criteria?
•New, Important and Broad
- creating new research topics;
- moving a field significantly forward;
- having impact on adjacent subfields;
- interesting to a broad readership.
Which format?
•4-page Letter format; strict length limit.
Preparing a paper for submission
How to write a good paper?
• What information is essential and necessary and should
be included in the paper?
– Big picture:
context and main result - novelty, importance, etc.
– Technical information and description.
• Overall organization:
– Abstract; Introduction; Main Body; Summary and Conclusion:
• Different parts, different emphasis;
• Balance: Big picture vs. Technical;
• Connections: A narrative - story.
How to write a good paper?
Abstract
•
•
•
•
Clear, brief
the scientific problem,
the main, new results,
importance, implications.
How to write a good paper?
Introduction
• The research problem and goal:
– scientific context (background and history) and motivation;
– objective, adequate referencing to the existing literature.
• The main results:
– Why new and why important?
– For PRL: why broad?
• Concise, informative, avoid jargon:
– For PRL: accessible, interesting to nonspecialists
How to write a good paper?
Main Body
• Necessary technical information:
– technical definition of the problem or experiment;
– how the theoretical problem is solved or
– how the experiment is executed;
– Not every detail, but self-contained: Allow other
researchers to reproduce your work!
• Logical presentation and discussion of the
results
How to write a good paper?
Summary and Conclusion
• Summarize the main results and
their physical meaning
• Discuss their implications for future
research or present a scientific
outlook
How to write a good paper?
• Style and Language
– Language: simple, direct, and meaningful;
– English: ask others for help if needed
(the editors cannot do this: we are sorry!)
– Spell checking and careful proof reading
• Technical Editing:
– Figure check: clarity and necessary
technical information
Useful resources for authors
(1) “Whitesides’ Group: Writing a Paper”, George M.
Whitesides, Advanced Materials 16, 1375 (2004)
A classic paper on how to write scientific papers that every researcher should read.
(2) “What Editors Want”, Lynn Worsham, The Chronicle of
Higher Education, September 8, 2008
http://chronicle.com/jobs/news/2008/09/2008090801c.htm
A journal editor reveals the most common mistakes academics make when they
submit manuscripts.
(3) Strunk and White, The Elements of Style (MacMillan:
New York 1979, 3rd ed.
So successful that it is known not by its title but as “The Little Book”.
Authors’ Participation in the Review Process
•
•
Be constructive and make progress
For resubmission:
–
point-to-point letter of response to all referee comments:
•
when agree with referees' criticism: provide response and
describe corresponding changes in the paper
•
when disagree: provide your counterarguments. Don’t be
confrontational!
–
A revised manuscript and an explicit list of changes made
–
Additional confidential comments to the editors if needed
How do the editors find referees for a paper?
We look for referees in:
• references (authors of, referees of)
• related papers in Web of Science, SPIN, NASA,
Google, APS database (authors, citing papers)
• suggested referees
• referee expertise in APS database
• mental database
We generally avoid:
• Undesirable referees
• Coauthors (current or previous)
• Referees at same institution as authors
• Acknowledged persons
• Direct competitors (if known)
• Busy referees (currently reviewing for PR/PRL)
• Overburdened referees (> 15 mss/past year)
• Consistently slow referees (>8 weeks to review)
• Referees who consistently provide poor reports
We have > 50,000 referees in the database,
but we are using “only” 1500 heavily(*)
Referee Statistics
(*) > 10 papers/year
Recognizing APS Referees
http://publish.aps.org/OutstandingReferees
2007
_________________________________________________
T. Ando
Tokyo Inst. of Technology
W. Bentz
Tokai University
Yutaka Moritomo University of Tsukuba
Kazuhito Ohmi
High Energy Accelerator Res. Org.(KEK)
K. Yazaki
Tokyo Woman's Christian University
Our database contains more than
50,000 referees. We consulted
25,000 referees last year.
Over 60 % are from outside the US.
But:
Country Papers Active
papers/
refs
PR/PRL Refs
Japan
2050
1485
1.4
China
3800
256
14.8
USA
7164 10650
0.7
World
30799 27316
1.1
Conclusion:
We need more participation
from Japanese referees!
The total number of referees in our database has
doubled in ten years
The PRL review process:
A schematic view
Manuscript
submission
Referees
consulted
Accepted
(~ 35%)
Rejected
(~ 65%)*
Editorial rejection
(~ 25%)
Decision letter resubmission?
* out of which
about ! are
published in PRA-E
The editors’ role:
to conduct an impartial & thorough scientific review
Editors are not technical experts (in general)
!
but they do strive to make sure that:
- no obvious conflicts of interest occur
- referees are experts in the field they review
- reports are detailed and substantiated
- response of authors is complete, dispassionate, and
substantiated
- review process is timely (*)
- review process is converging to a yes/no decision
- no special groups are favored/discriminated against
(*) this has many direct implications
Editorial constraints:
time vs. depth of review
11000 submissions to PRL in 2007
Staff: 17 in house; 5 remote (~12 full time editors)
• ~ 900 manuscripts / editor / year
• ~ 4 new manuscripts / editor / workday
• Max. average time per manuscript is ~2 hrs
More emphasis on fewer papers focus on quality assessment
• Fewer papers / editor
• Newly available software tools allow
better assessment of papers (in
choosing referees, cross-checking
facts, etc.)
- BUT take more
time
• Editors reject more papers based on
their own judgment (~24% in 2007),
thus freeing up time for the ‘better’
papers
Growth and Internationalization
Submissions to PRL
U.S.
Europe
Rest
33%
40%
27%
Published
Letters
2007
U.S. 35%
Europe
45%
PRL Evaluation Committee Report 2004
Committee Members
Report available at:
http://publish.aps.org/reports/PRLReportRev.pdf
PRL Evaluation Committee Report 2004
“A journal rises or falls on the quality of its refereeing.
Without talented, diligent referees interacting with first-rate
editors, a refereed journal becomes indistinguishable from
a preprint server.”
“[…] multiple reviews of bad and marginal papers continue
to demand too much editor time.”
“Weak or vague referee reports make it difficult to turn
down papers.”
“Papers frequently do not have the broad enough appeal
because they are aimed at narrow specialists. The
introductory paragraphs are rarely understandable by
non-specialists in the field.”
PRL Evaluation Committee Report 2004
Some subfields […] seem to regard small incremental
achievements as worthy of publication in PRL.
=============== Recommendations ===============
[The Committee] “encourage[s] the editorial staff to
[…] up to 20-25% of papers […
[…]
! reject […
based on internal review and assessment,
rather than sending them to external referees.
! Encourage author self-discipline in determining what to submit.
! Expand the editorial staff in underrepresented
geographical areas
Fast-track exceptional papers
PRL Evaluation Committee Report 2004
[Encourage] Referees to identify cutting-edge papers worth
publishing even if their correctness cannot be definitively
established. Referee training should emphasize that a
stronger attempt be made to accept more speculative
exciting papers that really move science forward.
! Have editors interact more closely
with the physics community
! Develop a pool of new younger referees armed with the
advice of experienced first-rate referees
! Take proactive steps to encourage the best referees
(“rewarding”
rewarding” referees)
Rejection without external review
Why it makes sense: result of a synergy of factors
2007: > 11,000 submissions to PRL
1 new PR/PRL paper / 3 minutes
Referee resources: increasing but limited
1 new referee / 30 minutes
+ it takes time to know and trust new referees
+ More journals, more papers, compete for referees’ time
Rejection without external review
PRL Review Committee:
“encourage[s] the editorial staff to reject […
[…] up to 20-25% of
papers […
[…] based on internal review and assessment”
2004: “Silent” experiment of pre-marking weak papers
2005: Experiment goes live
2008: As we gain experience, we ramp up rejections
without external review
Editorially rejected PRL: 5% chance of being published
Average PRL: 33%
Rejection without external review
Why not “rejection without peer review”?
Why, and to what extent, are editors “peers”?
================== Editors ==================
- PhD scientists
- most have postdoctoral experience
- some have held posts as professors
- some are active research scientists
- they read 900 Letters / year!
- they amass experience on their editorial decisions
(feedback, “reverse editing”, participation in APS
committees judging papers, etc.)
===========================================
=> Editors can often act as effective “peer-reviewers” as far
as broad interest & readability of paper is concerned
Rejection without external review
Red flags that may warrant editorial rejection
- Sloppy, opaque writing
- abstract and introduction too technical; nonunderstandable by non-specialists
- too much jargon / acronyms
- inadequate referencing
- too many old / specialized / self / ‘confined’ references
- no punch-line:
what is the main message of the Letter?
why is it important?
24.0%
21.3%
20.2%
10.9%
12.7%
12.6%
12.7%
~5% eventually published
compare: average PRL, ~33%
Why so many papers?
Are publication standards slipping?
Is the review process ethnically,
culturally, etc. biased?
Statistical analysis may offer some rationalization with
regard to these questions.
APS publications of most economies ~ 0.4 * GDP
(I.e. it takes about ~ 2.5 billion US$ to get one paper published)
(in billion US$)
Notes:
• Phenomenological relation, checked only for 2000-2005.
• 39 countries make up for 91% of World GDP and 97% of World APS publications
• GDP is based on the official exchange rates (OER), this is not the PPP-based GDP .
• For international collaborations, a manuscript is assigned to the country of its corresponding author.
Publications in the Phys. Review journals ~ GDP for most countries
AR: Argentina AT: Austria
AU: Australia
CZ: Czech Rep.DE: Germany DK: Denmark
HU: Hungary IE: Ireland
IN: India
NL: Netherlands
NO: Norway
TR: Turkey
UK: U. Kingdom
BE: Belgium
BR: Brazil
CA: Canada
ES: Spain
EU: Eur. Union FI: Finland
IL: Israel
IR: Iran
IT: Italy
NZ: N. Zealand PL: Poland
PT: Portugal
US: USA
ZA: South Africa
CH: Switzerland CL: Chile
FR: France
GR: Greece
JP: Japan
KR: S. Korea
RU: Russia
SE: Sweden
SOURCE: World Development Indicators database, World Bank: http://devdata.worldbank.org/data-query/
Publications in all Phys. Rev. journals vs. GDP
for 1995-2007 for the world’s richest economies
time
SOURCE: World Development Indicators database, World Bank: http://devdata.worldbank.org/data-query/
CN: China
HK:Hong Kong
MX: Mexico
SG: Singapore
• Internal metrics:
– Rejection Rate
PRL - receipts by country
PRL - receipts by country (US excluded)
PRL - publications by country
LATEST DEVELOPMENTS
• Exactly 50 years and 2 weeks after the
start of Physical Review Letters….
• APS is starting a new on-line Journal,
with the name of…..
Physics
Why are we doing this?
• We publish 18,000 articles each year
• Some articles are important, but can
only be understood by an expert
• We want to make the most important
new results accessible, with an
introduction by an expert scientist at a
level that all physics students and
researchers can understand
Our goal is:
“to highlight the best physics”
Physics will consist of:
• “Viewpoints”
–
written by experts about a paper recently published in PRL or PR and
describe to the non-expert why the paper is new, interesting, and
important (like News and Views, or Perspectives)
• “Trends”
–
written by experts and give an overview of a field, pointing to
connections and new directions
• “Synopses”
–
of selected papers in PR and PRL
• Links to “Focus” articles
http://physics.aps.org
Measuring citation impact
• Journal Impact Factor (JIF) measures
average citation performance of a journal
over a certain time-window…
but the number of citations varies widely
within any journal (by 2-3 orders of
magnitude)
• Much of the actual impact of a journal (or a
researcher, a university department, etc.) in
science is due to a small number of highly
successful papers: a modest subset of the
whole
Why the impact factor does not
say it all: It is an average.
Impact Factor = Average Citation Density
N
IF2006 =
citations2006
=
papers2004"5
# c(n)
1
N
The IF is the number of citations
over a 2-year window, averaged
over the whole journal.
The IF is the surface area of c(n),
normalized to the total number of papers N
!
Not all papers
are created equal!
Impact of top-cited portion of journals
1
PRL citation density ~
rank
(cumulative Impact Factor)
!
1
Power density
law is ‘universal’
citation
~
rank x
Journal
x
New J Phys
0.60
Science (ph) 0.54
!
Nat Mater
0.54
Opt Expr
0.52
Nature (ph)
0.51
PRA
0.48
PRL
0.47
Science
0.46
Nano Lett
0.46
APL
0.44
PRB
0.44
J-index: “Hirsch” index of papers
published in 2004-5, cited in 2006
75
44
40 40
31 31
23
27
20
18
13
Ranking physics journals by J-index
Journal
J-index
JIF
(significant
(average
performance)
performance)
_______________________________________________
Nature (ph)
44
22
Science (ph)
40
23.5
PRL
40
7
Nat Mater
31
19
Nano Lett
31
10
APL
27
3.9
PRB
23
3.1
RMP
23
33.5
Opt Expr
20
4
PRA
18
3
New J Phys
13
3.7
J. Phys. Soc. Jap. 12
2.2
Jap. J. Appl. Phys. 12
1.2
Progr. Theor. Phys. 7
1.9
Performance Indicators
average
significant
performance
performance
______________________________________________
Reseacher
citations/paper
H-index
Journal
JIF
J-index
Dept
“D”-index
Country
“C”-index
______________________________________________
Indicators of average vs. significant performance are
complementary:
Average Indicator: works best for small ! (e.g., JIF)
Significant Indicator: works best for large ! (e.g., J-index)
Citation conclusions
• Journal Impact Factor (JIF):
- average citation performance of journal
• Citation Density Curve:
- full citation performance of a journal
- upper bound for the JIF vs. the # papers published
- extract J-index and use to rank best journal
performance
• No physics journal that publishes:
> 500 papers/year can have a JIF > 20
> 150 papers/year can have a JIF > 33
(large impact factors are only possible for small journals)
Citation conclusions (cont’ed)
• Universal behavior: Citation density ~
1
rank
=> ALL journals have a few highly cited papers,
and many little-cited papers
!
=> it does not make sense to choose where to submit
your best work judging on impact factors!
(It only makes sense in an average sort of way, but
your best work is not an average paper)
Keeping Pace with the Growth of Physics
Published Letters
Raising the bar: Should we publish
fewer Letters?
(1) How do we attract and highlight the top papers…
…but also continue to provide broad coverage?
(2) Out of the ~4000 Letters published in 2007…
…how many were exceptional papers?
• Reported major discoveries?
• Moved an entire field?
• Created a whole new research area?
• Led to cross-field fertilization?
a. 100
b. 300
c. 500
d. 1000
(3) How many more Letters should we have published?
A. 20%
B. 50%
C. 100%
D. 200%
(4) What should we have done about the rest?
!"#$%!
Acknowledgments
Gene Sprouse (APS Editor-in-Chief)
Reiny Schuhmann (PRL Managing Editor)
Ling Miao (PRL), Sami Mitra (PRL), Dirk Jan Bukman (PRE)
Barbara Gill (APS)
... And to many other colleagues in the APS Editorial Office
Feedback / questions:
[email protected]