It is nearly impossible to determine which studies are bringing science forward now. But might it be possible to somehow operationalize this for studies written a generation ago, now that we understand their contribution much better? I guess quantifying remains hard, even if qualitatively it becomes easier going back in time. Might a ranking of papers work or comparing pairs of papers?

Then we could at least look whether in the past citations were correlated well with contribution to science. For my own publication list, I would say the correlation is weak, especially among the more cited papers.

Would it be an interesting study to have authors rank their own publication list and then compare that to citations? Authors may naturally be influenced in their ranking by the number of citations, which most will know to a certain degree.