Do Journal Rejection Rates Index Consensus?

Abstract

Hargens has made a significant contribution to our continuing attempt to understand the cognitive and social differences between scientific disciplines. A primary issue is to what extent does the level of cognitive consensus in the natural and the social sciences differ? Hargens (1975) and most others (e.g., Kuhn 1970, Lodahl and Gordon 1972, Price 1970, Storer 1967, and Zuckerman and Merton 1971; 1973) have long advocated the accepted assumption that the natural sciences have higher levels of consensus than the social sciences. We have argued that, at the frontier, where new knowledge is currently being produced (as opposed to the core where a small set of theories and analytic techniques represent the given at the time), all sciences have similar levels of relatively low consensus (S. Cole 1978; S. Cole, J.R. Cole, and Dietrich 1978; S. Cole 1979; S. Cole, J.R. Cole, and G. Simon 1981; and, primarily, S. Cole 1983). Hargens' article is an important new contribution to this debate. He focuses on journal rejection rates that are generally higher in the social sciences than in the natural sciences. He interprets these data as supporting the hypothesis that large differences exist in cognitive consensus, disconfirming Cole's (1983) theory. To stimulate further research on this topic, we raise several questions about the conclusions Hargens has drawn from his data and model. Hargens argues that the lower rejection rate in natural sciences journals is evidence that these fields have more cognitive consensus. We disagree: He believes that journal rejection rates are a better indicator of field differences in cognitive consensus than CR (defined formally below), the extent to which independent evaluators of a scientific product like a journal article or a research proposal reach the same conclusion. We believe that the latter measure is the best way to compare levels of consensus, and we view journal rejection rates as an inadequate indicator of consensus because they are influenced by consensus and other variables. Our comments are divided into two sections: the problems with journal rejection rates as an indicator and the merits of CR as an indicator. We also point out two problems in the formal model presented by Hargens.