[mgkuhn]

I thought the proper way to correct questionable results is to conduct and publish a follow-up study that independently looks at the same question with better data and better methodology. And wait until multiple independent teams have done the same. And then write a meta-analysis on the emerging pool of independent papers.

That's how scientific consent is normally formed, at least in rigorous disciplines like experimental physics or medicine. A single paper in the end is going to be just a single data point in any such meta-analysis study.

[fabian2k]

It depends on whether the paper is simply wrong for some reason, or if there is either fraud or a fundamental mistake in the procedure. But in general you're right, retractions are not the way to handle most scientific disputes.

Your point about consensus unfortunately doesn't quite work in cases like this were the people using the paper are not scientists. They're not continuing work in the same area, people are using this paper to support their arguments.