No Comments

There is the data, and then there is the interpretation of the data. They are not the same thing, although this line gets blurred too often.

Many years ago Andy Moiseff, a colleague of mine, showed me a ‘Letters to The Editor’, published in 1887 in Science. (G Hay. Instinct in the cochroach. Science Vol IX 1887  p. 622 ). I used to present this article as my first lecture to emphasise the importance of extracting the data from an article, and the perils of relying too heavily on the author’s description or interpretation.

In this letter Hay describes his experiment. His initial observation of the biological phenomenon is:

I have observed that a full-grown cockroach would climb up the gas-pipe and along the bracket towards the burner, but, finding the bracket a few inches from the flame too hot to traverse, would crawl back a few inches, wait a second or two, and then return towards the flame. If uninterfered with, he would, after a few trials, leave the bracket altogether, and return down the pipe, and run off at full speed.

His experiment is to wait for the cockroach to approach the flame, but now heat the other end of the bracket, essentially locking the cockroach between two hot spots. He describes the cockroach running between the two hotspots and, when not being able to escape, leaping onto the bench. And here is the beauty: he crushes the insect with his boot

because I wanted to observe the action of a fresh cockroach every time under the same circumstances.

Nice, simple experiment (remember, this is 1887). And those are the data. Now here is his interpretation of the data:

This is exactly what takes place when a fire occurs in a high building. The inmates (particularly women) jump wildly form the upper windows without waiting to see whether all other means of escape are exhausted — and get smashed on the pavement. Our friend ‘the unspeakable Turk’ says that women have no souls, and yet, although much higher in organisation than the cockroach, they act, in similar circumstances, precisely in the same way.

Oh what a jewel! This is such a beautiful example of how our cultural biases enter our scientific interpretation of data. We also interpret data based on the knowledge available at the time. Both of them continuously change. But the data remains the data.

All too often we dismiss whole papers based on what the author says about the data, and in the process run the risk of ignoring great pieces of experimental work.