In a letter to Science, William F. Perrin, a past editor of Marine Mammal Science and a present associate editor of the Journal of Mammalogy writes:
I have had great difficulty in lining up reviewers. Sometimes it takes 8 or 10 tries to find someone who will agree to review a paper. The typical excuse is “I’m too busy.”
First I try the people who have published the most relevant and recent papers on the topic in question. Then I move down the range of choices. The temptation, and sometimes the need, is to turn to potential reviewers in less-related fields or those who are not so “busy” (i.e., are not producing much themselves). This inevitably leads to less-knowledgeable reviewers and often reviews of lesser quality, which of course complicates the editor’s job and sometimes enrages the authors.
If an average acceptance rate of 50% is assumed, and if each paper needs at least two reviews, then each paper published represents at least four reviews. Following this logic, if you publish three or four papers a year, you should be doing at least 12 to 16 reviews. Anything less means that you are sloughing off the work to others who are perhaps less knowledgeable and capable than you in your specialty, and you should not be upset when someone reviewing a paper of yours “doesn’t have a clue.”
Doing a fair share of peer reviews should be a recognized and expected part of the job for scientific professionals; it should be written into the job descriptions of salaried scientists and be considered in evaluating junior faculty for tenure. The caution should be “Publish and review, or perish.”
The average of published papers in semantics and pragmatics may be closer to 2 or 3 a year. So, this calculation would mean 8 to 12 reviews per year. Are you doing that many?
We here at S&P have promised our editorial board members not to expect more than 2 reviews each year (although we maybe should be getting nervous about the load of manuscripts that we’re receiving). Assuming that they review for other journals as well, that probably does add up to the “ideal” workload for reviewing.