[davidp]

I have a friend who is a linguistics professor at a major public university, and she has complained to me about some of the projects she's been involved with where her fellow researchers really are not good at math.

Among other things, there's a lot of confusion about statistical significance, and how and when it's calculated. There are people out there, especially in the cognitive and social sciences, who think it's OK to run an experiment to gather some data, and then look for statistically significant trends in the data and draw whatever conclusions they can. They're p-hacking and they don't even know they're doing it; they think it's science. And the people judging their papers sometimes think the same thing. How do you fix a field like this?

This issue lies at the root of the reproducibility problem. The example someone gave about an undergrad cheering (that she didn't have to do any more math) points at one of the causes.