False theories die with disproof, but false data may live forever, or so my undergraduate advisor, Richard D. Alexander, told me. A single false fact can corrupt a dataset, a study, even a field. I remembered this as I counted the cells in wasp nest after wasp nest, patiently moving a dental mirror from one side to the other, nudging nighttime wasps aside so my cell count would be correct. I worked half the night in hot Brackenridge Field Laboratory meadows. After all, what would happen if it looked like there were fewer instead of more cells a few nights later?
Perhaps I need not have worried about those thousands of measurements, as accurate as I could make them with one, two, even three or four, counts on difficult nests. But I kept what Dick said in mind, ever aware of how devastating a false claim could be.
Fairly recently two claims of false data have come my way, and they are false themselves, though there are others that are real. The most recent one is on Stephen Jay Gould’s reanalysis of Samuel George Morton’s measures of skulls. Gould made a big deal of this, concluding that Morton’s racism caused him to mismeasure skulls. That this is not true is definitively shown by Lewis and colleagues in a recent PLoS Biology paper. I won’t reiterate the arguments, for then you might not click on the freely available paper. But they are completely convincing, relying on reanalysis, discussion of skulls and populations included and excluded, and even remeasurement of some skulls.
The second case involves Mendel’s pea proportions, which R. A. Fisher felt had numbers that were suspiciously good. Again, I’ll let the original authors speak for themselves. Novitski did a reanalysis published in Genetics in 2004, also subsequently discussed by Hartl and Fairbanks in 2007. The bottom line is that it is clear that Mendel reported things as he saw them, without fudging the data.
I hope you don’t now think I’m unconvinced anyone ever fudges data, for we all know that is not the case. In fact there is a whole blog based on looking for retractions. Check it out.
There is also a recent paper in Nature on some of the past individuals accused of wrong doing:
Clearly a false claim of wrong doing can be very damaging, but it is ground we must tread, though carefully. The more we put our datasets up for all to reanalyze, the more we will be able to verify analyses. Full replications of studies remain essential, though they are all-too-rare in behavior, and failure-to-replicate can be from different conditions, not falsification.
Science is a human endeavor, with much potential for error. It is our job to avoid error, expose falsification, and understand its devastating impact.