“It was a great fight and I don't think there was anything wrong,” announced racer Lewis Hamilton, who was accused of cutting a corner at the Belgian Formula One Grand Prix. As I watched, I mused about corner-cutting in science, and whether such practices are justified or even necessary in order to succeed.

When data are presented, the reader or listener assumes they are robustly reproducible. One trusts that quantitative results are based on an adequate number of experimental replicates and reproducible results, and that the design includes appropriate controls. Are such assumptions necessarily valid? Much may be left unsaid, especially in a culture in which it is important to save face.

A student from another lab once sought my advice on alternative experimental approaches, claiming that her original one had failed. I later discovered that she had attempted the experiment only once, and without proper controls. Even in the collegial atmosphere of lab meetings, there is pressure to look good in front of both peers and supervisor. The emphasis on positive data is quite strong. Negative data, technical problems and methodological shortcomings may be overlooked.

Hamilton was penalized for his alleged corner-cutting. But short cuts in the lab may never be detected — even though they could matter a great deal.