If they really answered a question that comprehensively, then wouldn't they be cited everywhere as the paper that established the definitive answer in the field? I think by definition papers could not be both the true answer and poorly cited.

For example, Nature published recently a paper by Denver et al. that established a mutation rate for C. elegans. It will not completely preclude people from continuing to work in the area, but it does give An Answer, and is therefore cited already many places as The Mutation Rate in worms.

Shall we nominate Chris' comment as our first category killer?

Actually, large-scale negative association studies spring to mind as the quintessential category killer. The best of these studies provide a definitive negative answer that stands as the final word in the field. And, perhaps not surprisingly, they are relatively poorly cited, inspiring only modest follow-up and discussion as the field quickly moves on to consider new, more promising candidates or dissect the few positive results that withstand the battering and close scrutiny of independent replication.

A case in point is a paper we published in October 2003 describing a lack of association between germline variation in MSR1 and prostate cancer risk. As of this morning, the paper has been cited 10 times, ranking last among all the papers published in that issue. (The "average" paper in that issue has been cited 43 times.)

This isn't to say, of course, that negative association results aren't important for the field. Indeed, they're essential, as Mark Patterson and Lon Cardon have recently argued. But despite their value, they're unlikely to become citation classics. Mini category killers, as it were.

Concerning human genetics and possible category killers I would like to share one of my thoughts. In April 2005, several teams reported that a Complement Factor H Variant increases the risk of age-related macular degeneration (AMD). Subsequently other teams around the world i.e US (IA and MI) Germany and Iceland independently confirmed this finding. A large population risk attributable to this variant was estimated , PAR ~40%. At that time I thought that this massive finding might have discouraged people from looking for another major genetic player in AMD, even if it is obvious that in complex disease more than one factor is involved. On the contrary, interest in this area still seems high and has led to 2 publications of other loci around LOC387715 on 10q26 by association study (also with large PAR). One can argue that these papers were probably submitted prior to the publication of the CFH papers and did not have time to kill enthusiasm. At least it means that it is difficult to know when a question is fully answered in complex diseases, since multiple loci are likely to act together and/or that interpopulation heterogeneity exists. In any case, good luck to those who try to comprehend and summarize today's literature, since a single individual lacks the time to read everything.