The Love‐at‐First‐Sight Effect in Research

PROMISING results and the favorable outcome of an investigation lead naturally to a euphoric response, particularly when they exceed expectations. Frequently, however, other investigators (or sometimes the same investigator) will attempt to reproduce these results, and they see none of the nice things that were first seen. Often, the second time around, slightly different materials or methods are used, or we find the first report was a misreport. But more often, the early enthusiasm is the result of an artifact which reasonable repetition cancels out, and reasonable statistical precaution might have hinted at. There are even some situations in which a repeat must give poorer results than the first try. The reason is obvious once it is seen-and yet people have devoted much experimental work to trying to find out why a material "deteriorated"-when the putative deterioration was a pure artifact. Three examples, derived from widely different research areas, illustrate the problem of the sometimes misleading nature of early work. In the first two, a relatively homogeneous material, based on electrocardiograms of "normal" subjects was chosen.' The researcher (call him A) attempted a correlation between the subjects' ages and the various ECG measurements. The spatial magnitude of the maximal QRS vector was selected as the first candidate measurement. There was little reason to believe that the two variables were related, other than indirectly, and a low correlation was expected. Using 25 records from subjects ranging in age from 20 to 85 years, A found a large negative correlation of -0.63. He was careful, and he looked up the 95% confidence limits on r -0.63, N 25 (the Geigy tables2 are an excellent source), and found them to be -0.31 to -0.82. This outcome was statistically significant, new, surprising and, as might have