Hacker News new | past | comments | ask | show | jobs | submit login

> lately it feels like _Science_ and _Nature_ are in fact especially likely to collect results that are very exciting and also wrong, because they feel like they're the best journals so their results should be surprising.

I believe this is accurate on all counts.

Slate Star Codex put the issue fairly pithily: there's only one way to get surprising-if-true results that isn't surprising.

Here's how I explain it: Science and Nature play a role different from other journals. Their role is to stimulate new thoughts amongst the most skilled in the field. Any really good scientist in a field can read a paper published by another really good scientist and hone away the bullshit, but be inspired by the ideas. People who are not skilled in the art of reading a competitive paper are going to be disappointed by the sloppy lab work, sloppy writing, sloppy thinking, because it appears that the overall benefit of disseminating new ideas widely for experts to cogitate on may exceed the value from having "perfectly correct" papers.

Case in point: the original DNA structure paper (W&C, 1953) actually had a wrong detailed structure for the DNA (but the right conclusion about how the structure provided a hypothesis for a templating mechanism) because the alcohol concentration used to form the crystal was too high. The value of getting the paper out the community ASAP greatly exceeded the small details that were wrong (which eventually got resolved), because it stimulated the thought of the community around the templating mechanism.

Can you explain the joke?

The only way to get a surprising-if-true result that is not surprising is if the result is not true. Hence the joke is that most of these surprising findings are probably B.S.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact