CORRELATION AND CAUSALITY

"In your October 1985 article on Hormesis," writes reader Lloyd McAulay of New York City, "you refer to evidence based [only] on correlation... I am not sure that I see the distinction between correlation and causality except as to degree. It seems to me that all evidence is correlation type evidence..."

Let me give some drastic examples of correlations that clearly do not indicate a cause-effect law, but merely a relationship between two variables. The number of hairs on my head (first variable, x) is steadily decreasing; the taxes I pay (second variable, y) are steadily rising, so that I can plot a curve giving a clear relationship between the two numbers: it is unambiguous, and the points lie unmistakably on a definite curve. I doubt that it is the tax increases that make my hair fall out, and conversely, it is not the hairs on my head that the income-redistributors use for a tax policy. That curve is nothing but a way of depicting the history of taxation using my hair as a calendar or clock.

Now take a statistical case: the more traffic lights in a city (x), the higher the incidence of eye disease among its drivers (y). If you doubt my medical erudition, just plot a point for each US city as given by these x, y coordinates. You will soon see that what you are plotting is essentially the size of the cities: Chicago has a lot of traffic lights and eye patients, whereas Jackson, Wyoming, has little of either. The reason why the points do not lie on a line or curve, but form a cloud about a mean fine is that the relationship is statistical: the size of a city is not the only factor determining the number of eye patients (or traffic lights).

Next, take a case where things are less transparent. Record the number of cars going past your house (x), and the distance (y) to the furthest radio station that has just appeared on, or disappeared from, your AM dial. You should again get a statistical cloud, somewhat more scattered than above, but clearly hovering about a mean line that says the two vary up and down together.

Cause and effect? Yes, but not of each other: both are results of the same cause, namely the sun, which ultimately determines our daily schedule and rush hours (cars), and also causes the diurnal variation of ionized layers high in the atmosphere that reflect and attenuate radio waves (distance to station).

I don't think the physicians know for sure, and certainly I don't.