No one can become a genuine scientist without doing practical work, as John Baruch points out (see Nature 507, 141; 2014) — whether it is at the lab bench or desk. But it is not the successful experiments that count so much as the frustrations inflicted by those that fail.

It is the embarrassment of a statistical analysis that reveals inadvertent data-input errors; it is tearing one's hair out debugging a program that once worked well; it is learning that “nature cannot be fooled” (as physicist Richard Feynman warned).

This experience of failure and error is unique to science. In the humanities, a student learns from mistakes such as not remembering to use specialist vocabulary, failing to put forward established arguments or not engaging in self-promotion before and after seminars. Practical science demonstrates that doing science involves more than just picking up such tricks.