Causal relationship between article citedness and journal impact

The relationship between article citedness and journal impact was investigated on the basis of complete publication lists provided by 16 senior scientists from a major Norwegian biomedical research institute. The citedness of each individual journal article was measured as the mean annual number of citations during the second to fourth year after publication, and compared with the impact factor (mean article citedness) of each corresponding journal, recorded during the first two years after publication. The distribution of article citedness was very skewed (corresponding to a topnormal, extreme‐property distribution yielding linearity in a semilogarithmic plot), even for individual authors and within defined journal impact cohorts. The large variability was not due to random citation of the individual article, but rather to a fundamental heterogeneity among the articles. A similar skewness was found in the distribution of journal impact values, whereas the author impact distribution was less heterogeneous. The skewed distributions resulted in poor correlations between article citedness and journal impact, for the whole article population (correlation coefficient, c=0.41) as well as for individual authors (mean c=0.32; range 0.05–0.66) and journal pairs (mean c=0.22). Pooling of articles into defined journal impact cohorts dramatically improved the correlation (c=0.999 for the study group as a whole), indicating that the authors tended to submit their most cited work to journals of higher impact (although several notable exceptions were observed among individual authors). However, very large numbers of articles (50–100) had to be pooled in order to obtain good correlations (c=0.8–0.9). Furthermore, the mean citedness of the study group (3.34 ± 0.16 citations/article/year; mean ± SE of 907 articles) differed significantly from the corresponding mean journal impact (2.61 ± 0.07); for individual authors even larger discrepancies were observed. Use of journal impact as an evaluation parameter may therefore yield highly misleading results, unless the evaluated unit (author, research group, institution, or country) happens to be equal to the world average. By dividing the authors into a highly cited group (the eight most cited authors) and a less cited group (the eight least cited authors) it was possible to observe a twofold ratio in citedness between the two groups throughout the journal impact range. This difference could not be accounted for by journal choice, and did not diminish with increasing journal impact. The citedness of journal articles thus does not seem to be detectably influenced by the status of the journal in which they are published. © 1994 John Wiley & Sons, Inc.