I can't speak to all science, but I am a CS (cryptography) researcher and routinely read papers from the 1980s. The level of rigor and quality of my field's papers has absolutely improved, by leaps and bounds. The formal definitions in those early papers are often non-existent (and sometimes wrong in retrospect) and the proposed constructions are often much simpler (and sometimes subtly wrong in retrospect.) And the number of papers has increased by at least an order of magnitude.

On the flip side, those early papers contain the most fundamental discoveries in our field: you're only going to invent RSA or blind signatures or zero knowledge once. It's possible all those researchers were much smarter than we are now. (I grant this!) But there are a lot of absolutely brilliant people I know today. Alternatively, the lower-hanging fruit is all gone and the problems have become much harder.