In “Testing Relativity from the 1919 Eclipse—A Question of Bias” (Physics Today, March 2009, page 37), Daniel Kennefick makes synonyms of two words—observation and experiment—that traditionally have described different ways to gain knowledge of the physical world.

In his first paragraph, Kennefick describes Arthur Eddington’s work on the 1919 eclipse as an observation, but elsewhere he usually calls it an experiment. He mentions “observation” or some form of it 5 times in that first paragraph and does not mention “experiment.” Yet “experiment” or forms of it appear 19 times elsewhere in the text of the article where “observation” or forms of it appear 3 times.

In December 1919 Eddington wrote in the preface to the second edition of his Report on the Relativity Theory of Gravitation (Fleetway Press, 1920), “I think it may now be stated that Einstein’s law of gravitation is definitely established by observation.” Eddington appears never to have used the word “experiment” to describe results of the 1919 eclipse expedition, but he does use it to describe anticipated work on Fraunhofer lines, which agrees with traditional understanding of the word. And Albert Einstein himself, in early October 1919, reported that he had received provisional eclipse results of the “Beobachtung” (observation). 1  

Having spent 18 years doing experiments on water waves, I am aware, along with Kennefick, that measurement and insight—seeing the link between existing knowledge and the observed phenomenon—are forms of observation accompanying experiment. However, experiment differs essentially from observation by prescribing the values of variables believed to be relevant.

Although ongoing changes in dictionary definitions tend slightly to agree with Kennefick’s usages, I think the traditional distinction between observation and experiment is logically necessary for the results of the 1919 eclipse expedition. The data obtained then are observations, as Eddington and his contemporaries called them.

The Collected Papers of Albert Einstein
, vol.
 et al, eds.,
Princeton U. Press
Princeton, NJ
), p.