Science has long since been the dominant way of knowing. Its funding eclipses (by far) that of the humanities, generating a slew of research papers – 35 million for biomedical sciences alone, with a further 1 million added in this field every year. But all is not well: its reputation as the source of useful knowledge has been compromised by a marked decline over recent decades in novel insights and ‘breakthrough’ discoveries.

There are several reasons why this might be so, but the most important, as suggested by Csaba Szabo’s title, is that much scientific research nowadays is ‘unreliable’. The hallmark of reliability, he observes, is that genuine scientific findings should be replicable by others – the same experimental methods should produce the same results. In practice, this is assumed to be the case at publication stage, though a pharmaceutical company (for example) will need further confirmation that an experimental finding is ‘sound’ before investigating it in their drug development programme. 

Szabo’s book opens with the salutary experience of the biotechnology company Amgen, just over a decade ago, which first drew attention to the issue of reliability. In a brief, if seismic, commentary in the journal Nature, Amgen’s director described the outcome of seeking to replicate the results presented in fifty-three