by Michael Ernst (mernst@cs.washington.edu)
December, 2006
Contents:
- Why to prefer a conference
- Why to prefer a journal
- Other resources
(Also see my advice on writing atechnical paper, and a letter to USCIS making the point thatin computer science, papers inpeer-reviewed conferences are accepted as high-quality scholarlyarticles.)
(Note: This webpage is oriented toward computer scientists. Theinformation is not necessarily accurate for other scientific fields.)
Should you publish your work in a conference or in a journal? Each isappropriate in certain circ*mstances. This webpage lays out some of thetradeoffs.
This information is accurate to the best of my knowledge, and it agreeswith what all top researchers say. However, it is not intended to impugnany particular conference or journal — there are always exceptions toa rule.
Why to prefer a conference
In computer science, your preference should be for conferencepublication. Here are some reasons.
- Conferences have higher status. In part this is a historical artifact ofthe field of computer science, but it is self-perpetuating since that makesthe best researchers want to send their papers to conferences rather thanjournals.
- Conferences provide higher visibility and greater impact. Many people willattend your talk, you will have the opportunity to answer questions, andpeople will talk to both you and to one another in the hallways. Evendisregarding the event itself, more non-attendees read conferenceproceedings than read journals.
- Conferences have higher quality. Acceptance rates to good conferences areoften around 10% (at least in software engineering, which is my field),whereas even the best journals are less selective. Naturally, there existlow-quality conferences (and journals), but if your c.v. is cluttered withthem, then you will appear to be incapable of good work (even if the workyou published in those venues really is good!), and your good publicationswill not stand out. A good rule of thumb is that the best conferences aresponsored by ACM.
- Conferences are more timely. It can take years for a journal publicationto appear (or even for reviews to come back), whereas the turnaround timefor conference reviews is a few months, and the proceedings also appearquickly.
- Conferences have higher standards of novelty. Journals often only require20-30% of the material to be new, compared to an earlier conference version.
Why to prefer a journal
There are situations in which journal publication is desirable.
- Journals may have longer page limits. If you have too many experimentalresults to fit in a conference publication, then a journal affords anopportunity to include them. You can also include proofs that are too long(or boring) for a shorter publication. A journal paper could recap orgiven an overview of an entire research area.
- Journal reviews may be more detailed. Ideally, a journal reviewer willspend days on a paper, whereas a conference reviewer cannot afford to do sofor each of the many papers he or she is assigned. Conference reviewersmay believe the authors' claims (regarding a proof, for example), whereasjournal reviewers are expected to verify them. Especially in certainfields, reviewers may expect that a paper will be submitted (orre-submitted) to a journal. In any event, if you get a longer review, thatcan help you to improve your work or to understand its shortcomings.
- Journals give the opportunity to revise your work and re-submit it forreview. Actually, conferences give this too: if a paper is rejected fromone conference, then you can revise based on the reviewers' comments andsubmit to a different conference, or the same one the next year.
- Journals have higher acceptance rates, giving the opportunity to get yourresearch published. The same is true of workshops. These are particularlygood venues for people who are just starting their research careers.
- Some lesser-ranked universities evaluate faculty on the basis ofjournal publications, because the Dean of Engineering is unable or unwilling tounderstand computer science. In most scientific fields, journals havehigher standards than conferences; computer science is a rare exception.A top-ranked CS department can convince the dean to use the proper evaluation metric. A lower-ranked CSdepartment cannot (the dean may think the department is trying to fool himor her).If you are at one of these universities, you will need to publish injournals, probably by submitting slightly revised versions of yourconference papers to journals.The rush for people at lower-ranked universities (some of whom areexcellent researchers, and some of whom are not) to submit even marginalresults to journals is another regrettable factor that tends to lower theoverall quality of journals.
The best papers at a conference are often solicited for expedited journalpublication.I sometimes decline these opportunities, but your circ*mstances may be different.Whether you accept this invitation should be based on thefactors above, such as whether there is value to the community of anexpanded version of the paper, and how much more work it is to prepare thejournal version. (For example, is there a thesis, technical report, orother document with additional material beyond the conference paper? Evenbetter, are there additions that were suggested by reviewers or duringdiscussions at the conference?)
The journal version of a publication will be cited more than the conferenceversion, because the journal version has a later date and thus seems moreauthoritative. This is a good thing if the journal version adds real value(or corrects problems!). However, if you have cluttered the paper with alot of details that aren't crucial (like extra tables of results,experiments that support your point slightly less strongly than the mainones, or discussions of tangential issues), then your paper may actuallyhave less impact because readers will get mired in the irrelevant details.Good writing can avoid such problems.
Other resources
Regarding the impact of conferences vs. journals, see the CRABest Practices Memo,Evaluating Computer Scientists and Engineers For Promotion and Tenure and Bertrand Meyeret al.'s CACM articleResearchEvaluation for Computer Science.
My biographyindicates my qualifications to have an opinion on publication venues.
A Chinese translation of this webpage is available, thanks to Hao-Chuan Wang.
Back to Advice compiled by Michael Ernst.
Michael Ernst