THIS VIEW OF LIFE

Cordelia's Dilemma

Silence, though usually undervalued, can be golden

by Stephen Jay Gould

While Goneril and Regan jockey for their father's wealth by proclaiming their love for him in false and fulsome tones, Lear's third daughter, Cordelia, fears the accounting that her father will soon demand: "What shall Cordelia do? Love, and be silent...since, I am sure, my love's more ponderous than my tongue."

Lear then forces Cordelia into this game of ever more elaborate professions of love: "What can you say to draw a third more opulent than your sisters?" When the honorable Cordelia, refusing to play falsely for gain, says nothing, Lear cuts her off from all inheritance, proclaiming that "nothing will come of nothing."

Lear's tragic error, which shall lead to blinding, madness, and death, lies in not recognizing that silence--overt nothing--can embody the deepest and most important meaning of all. What, in all our history and literature, has been more eloquent than the silence of Jesus before Pilate, or Saint Thomas More's date with the headsman because he acknowledged that fealty forbade criticism of Henry VIII's marriage to Anne Boleyn, but maintained, literally to the death, his right to remain silent and not to approve?

The importance of negative results--nature's apparent silence or nonacquiescence to our expectations--is also a major concern in science. Of course, scientists acknowledge the vitality of a negative outcome and often try to generate such a result actively--as in trying to disprove a colleague's favored hypothesis. But the prevalence of negative results does pose an enormous, and largely unaddressed, problem in the reporting of scientific information. I do not speak of fraud, cover-up, finagling, or any other manifestation of pathological science (although such phenomena exist at a frequency that, in all honesty, we just do not know). I refer, rather, to the all too wonderfully human love of a good tale--and to our simple and utterly reasonable tendency to shun the inconclusive and the boring.

The great bulk of daily scientific work never sees the light of a published day (and who would wish for changes here, as the ever-increasing glut of journals makes keeping up in one's own field impossible and exploration of others inconceivable?). Truly false starts are deposited in circular files--fair enough. But experiments fully carried forth and leading to negative results end up, all too often, unpublished in manila folders within steel-drawer flies, known only to those who did the work and quickly forgotten even by them. We all know that thousands of novels, considered substandard by their authors, lie in drawers throughout the world. Do we also understand that even more experiments with negative results fill scientific cabinets?

Positive results, on the other hand, tell interesting stories and are usually written up for publication. Consequently, the available literature may present a strongly biased impression of efficacy and achieved understanding. Such biases, produced by the underreporting of negative results, do not only permeate the arcana and abstractions of academic science. Serious, even tragic, practical consequences often ensue. For example, spectacular medical claims for the efficacy of certain treatments (particularly for chronic and fatal illnesses like cancer and AIDS) may be promulgated after a single positive result (often obtained in a study based upon a very small sample). Later and larger studies may all fail to duplicate the positive results, effectively disproving the value of the treatment. But these subsequent negative results often appear only in highly technical journals read by more restricted audiences and, as nonstories, do not so readily attract the attention of the media--and people may continue to squander hope and waste precious time following useless procedures.

Statistics often get a bum rap in our epithets and editorials. But I am both a champion and a frequent user of statistical procedures, for the science exists largely to identify and root out hopes and misperceptions falsely read into numerical data. Statistics can tell us when published numbers truly point to the probability of a negative result, even though we, in our hopes, have mistakenly conferred a positive interpretation. But statistics cannot rescue us when we hide our nonlights under a bushel (with apologies to Matthew 5:15)--that is, when we only publish positive results and consign our probable negativities to nonscrutiny in our file drawers.

I had thought about this problem a great deal (especially when writing The Mis-measure of Man), but I had not realized that this special sort of bias had both a name and a small literature devoted to its weighty problems, until I came upon a paper by Colin B. Begg and Jesse A. Berlin entitled "Publication bias: a problem in interpreting medical data" (Journal of the Royal Statistical Society, vol. 151, 1988, pp. 419-63).

Begg and Berlin begin their paper with a wonderful quotation from Sir Francis

10   NATURAL HISTORY 2/93


185