> no non-scientist can evaluate the claims of [...] To be more precise: No non-expert in field X can evaluate claims with respect to field X. The phenomenon

> no non-scientist can evaluate the claims of [...] To be more precise: No non-expert in field X can evaluate claims with respect to field X. The phenomenon you're talking about is with respect to lay people. But there is a parallel phenomenon that happens with intelligent people with expertise in unrelated fields. It's relatively easy to read a handful of papers in a field other than your own and convince yourself that you are qualified to pass judgement. But how can you be sure you'd see a problem in the methodology of a paper, even if it existed? And the larger problem: how do you know the set of papers you read is representative? In the worst case, there could have been a shift in the field that no one bothered to document (perhaps because the shift was due to a lack of publishable results). This is my number one gripe reading old papers in computer science: sometimes an idea just disappears from the literature because frankly it didn't work, but you can't cite a paper that just says that outright, even though everyone (who works closely in the field) knows it to be true. In my experience, both problems take time and experience to solve. Which means that if you don't have those, you're better off finding an expert you trust and asking them---in person---what papers are worth reading.