- Small sample sizes (O(100))
- Low correlation coefficients
- Everybody is predisposed to believe the conclusion (what kind of monster doesn't like nature?)
- There is zero incentive to report a negative finding
And even if some of the studies do replicate, the response variable is often either self-reported, or performance on some written test or survey: so it's not really clear what we are even measuring. What is the unit of "general connectedness" ? How about "eudaimonic well-being" ? These are not legitimate scientific concepts.
Nature is great. I find that a long walk in the woods is a good way to clear my head. But I'm not convinced that these studies add anything to my common sense intuitions about the matter. And by using weak methods to justify something we all already believe (or want to believe), we risk degrading trust in science that will be greatly needed when scientists reach a conclusion that isn't politically convenient.