The skewness of science
Article first published online: 4 JAN 1999
Copyright © 1992 John Wiley & Sons, Inc.
Journal of the American Society for Information Science
Volume 43, Issue 9, pages 628–638, October 1992
How to Cite
Seglen, P. O. (1992), The skewness of science. J. Am. Soc. Inf. Sci., 43: 628–638. doi: 10.1002/(SICI)1097-4571(199210)43:9<628::AID-ASI5>3.0.CO;2-0
- Issue published online: 4 JAN 1999
- Article first published online: 4 JAN 1999
- Manuscript Revised: 25 MAY 1992
- Manuscript Accepted: 25 MAY 1992
- Manuscript Received: 16 MAR 1992
Scientific publications are cited to a variable extent. Distributions of article citedness are therefore found to be very skewed even for articles written by the same author, approaching linearity in a semilog plot. It is suggested that this pattern reflects a basic probability distribution with some similarity to the upper part of a normal (Gaussian) distribution. Such a distribution would be expected for various kinds of highly specialized human activity, parallels being found in the distribution of performance by top athletes and in the publication activity of university scientists. A similar skewness in the distribution of mean citedness of different authors may combine with the variability in citedness of each author's articles to form a two-leveled citational hierarchy. Such a model would be capable of accounting for the extremely skewed distribution of citedness observed for all articles within a scientific field, which approaches linearity in a double-log rather than in a semilog plot.
The skewness implies that there will always be a large fraction of uncited publications, the size of the fraction depending on the citation practices (such as the number of references per publication) within the field in question. However, as part of a continuous probability distribution even uncited articles have a definite probability of contributing to scientific progress. Since it is furthermore impossible to eliminate uncited articles for statistical reasons, they should be the cause of neither worry nor remedy.
The citational variability between articles in a journal is less (semilog linearity) than in the corresponding field as a whole, suggesting that each journal represents a select, stratified sample of the field. However, the variability is still too large to make the journal impact factor (the average citedness of the journal's articles) suitable as a parameter for evaluation of science. Fifteen percent of a journal's articles collect 50% of the citations, and the most cited half of the articles account for nearly 90% of the citations. Awarding the same value to all articles would therefore tend to conceal rather than to bring out differences between the contributing authors.
The skewness in the citedness distribution of each author's articles, the large overlap between different authors and the existence of field-dependent systematic differences in citedness would seem to make even article citations unsuitable for evaluation of individual scientists or research groups. At the national level, citations may be more useful, provided due corrections are made for the field effects. © 1992 John Wiley & Sons, Inc.