Publish-or-Perish Postscripts

Let me offer an alternative engineering perspective to Mohamed Gad-el-Hak's well-crafted Opinion piece (Physics Today, March 2004, page 61). Some journal articles properly target applied physicists- a group of industrial scholars who are typically experimentalists. The American Institute of Physics (AIP) and its member societies publish journals-for example, the Journal of Applied Physics, Review of Scientific Instruments, and the Journal of Rheology—that in part target that same important readership. The industrial physics audience rarely publishes; instead, it is composed of practitioners, who uncover clever ways of making money by applying physics. Their noble work secures and generates jobs and prosperity.

Some of us devote our research publications to such physicists. We craft papers to help them, to change the way they think, to affect how things are manufactured. Our search is narrower than the search for truth; we seek something more elusive: usefulness to the applied physics community. In my opinion, the real tragedy of the publish-or-perish phenomenon is not how rarely scholars cite one another. It is the precipitous decline in the proportion of papers of interest to applied physicists. Indeed, a journal's impact factor may be inversely proportional to its impact on the applied physics community.

Thus, only with the greatest care should journal impact factors and individual citation rates be included in analyses of tenure and promotion. Specifically, for faculty members who bravely seek usefulness to readers from industry, journal impact factors and individual citation rates lack relevance.

I am grateful for Mohamed Gad-el-Hak's thoughtful Opinion piece and his response to my suggestion (Physics Today, September 2004, page 11). However, I dispute his claim that none of the suggestions offered by readers directly address the problem of excess. My suggestion to reward referees would do just that for several reasons.

• Proper recognition of reviewing efforts would encourage each referee to be more thorough and thereby increase the likelihood of spotting and rejecting poor submissions.
• Referee recognition would also encourage journal publishers to subject each paper to a greater number of referees, because more would be willing to make themselves available. The net effect would again be to increase the opportunities for identifying poor submissions.
• The larger pool of referee volunteers would inevitably result in reduced paper output, since referees engaged in reviewing submissions have less time, and somewhat reduced motivation, to write papers of their own.

I respect Gad-el-Hak's reluctance to accept this idea because referee anonymity could be lost. Several work-arounds spring to mind: Instead of printing the names of referees, journals could provide a standardized receipt or certificate of gratitude for each review. Alternatively, if it was decided that an authorized automatic search should be able to find all articles refereed by a particular person, a protected automatic search system could be designed wherein the names of referees would be published in encrypted form in the paper they have reviewed; keys to the encryption would then be delivered only to the referees, who could verifiably demonstrate their refereeing contributions to those parties they choose.

Mohamed Gad-el-Hak's Opinion piece is one of several articles extolling the maintenance of high standards in research publications in the face of the increasing publish-or-perish frenzy in modern-day academia.¹

Gad-el-Hak wrote that "the number of citations per publication is a fairer index of competence than the total number of citations." Yet, in his Opinion piece and in the Letters that followed, there was no suggestion of any measure to address multi-author papers or the increased number of citations as a paper ages. I recently introduced the concept of the author impact factor (AIF) as a more precise measure of the impact of an individual author's published work.

In spring 2004, a committee in our department at the University of Florida suggested using the total number of citations (TC), along with the total number of papers, amount of funding, and other factors, as a measure for evaluating faculty performance for a specific purpose. I suggested a combination of the AIF, the TC, and the author's average journal impact factor (JIF) as a better way to evaluate a researcher's publication record.

Because the TC depends on various factors-among them age of papers, number of papers, and number of authors-it would be better to use additional measures. I defined the AIF as the average of the equivalent single-author annual citation rate (ESAACR) of all of an author's papers. For each paper, the ESAACR is the number of citations (NC) divided by the number of authors (NA) and the age of the paper (AP). That is,

Averaging the ESAACR of all papers published by an author gives that author's AIF. For example, a particular paper published in the Journal of Applied Mechanics in 1986 has the following data as of June 2004: NC = 145 citations, NA = 2 authors, AP = 18 years, and ESAACR = 4.028 citations a year for each author.

The AIF is a normalization not unlike what we do in science and engineering: When we talk about weight, we introduce the concept of mass (normalization with respect to acceleration of gravity) and density (normalization with respect to volume); when we talk about distance, we also consider speed (normalization with respect to time).

The concept of AIF is restricted to a particular field. To allow a comparison across fields, such as mechanical engineering and chemistry, we can consider the relative ESAACR, which is defined by dividing the ESAACR of a paper by the recent or current JIF of the journal in which that paper was published. For example, the above-mentioned paper in the Journal of Applied Mechanics, which had a JIF of 0.628 in 2002, has an ESAACR of 4.028 and a relative ESAACR of 6.414. Then the relative AIF is obtained by averaging the relative ESAACR over all papers published by a given author, using the JIFs in the same recent or current year for all journals in which that author's publications appeared.

An author with a relative AIF of 1 would have equivalent single-author publications with impact, on average, close to the JIFs of the peer-group journals in her field. On the other hand, because the AIF and the relative AIF are measures related to a single author, her own papers would actually have an impact higher than the JIFs of her peer-group journals, since papers are generally written by at least two authors.

The computation of the AIF and relative AIF can be done easily with any spreadsheet software, and could even be provided as an additional feature in the ISI Web of Science webpage at http://www.isinet.com.

Surely, the AIF and relative AIF would speak volumes about the publication impact of the dean that Gad-el-Hak mentioned, the one who published one paper a week.