Depending on Numbers
Theodore Porter
Distinguished Professor of History and Peter Reill Chair in European History
University of California, Los Angeles
Reproducibility in modern science is often defined in terms of probability and error. These are reasonably straightforward when it is a question of measuring an identical quantity according to a well-defined protocol. But reproducibility is more typically about finding new ways to measure, and often about objects subject to variation, such as living organisms. Even physical science works much harder to extend its findings than to repeat them, while in medicine and psychology, few results can be expected to replicate with high precision. Under such circumstances, reproducibility is not a rigorous standard, but more often a loose one that depends on expertise and interpretation. Moreover, science that matters will often be caught up in human interests. The effort to repress these by means of rigorous quantification succeeds only some of the time, and can easily create more problems than it solves. The perspective of history offers ways of understanding this dynamic, and even perhaps of adapting the ideal of reproducibility to the demands of practical and public science.