From Wikipedia, the free encyclopedia

The New York Times of November 10, 1919, reported on Einstein's confirmed prediction of gravitation on space, called the gravitational lens effect.

The predictive power of a scientific theory refers to its ability to generate testable predictions. Theories with strong predictive power are highly valued, because the predictions can often encourage the falsification of the theory. The concept of predictive power differs from explanatory or descriptive power (where phenomena that are already known are retrospectively explained by a given theory) in that it allows a prospective test of theoretical understanding.

Scientific ideas that do not confer any predictive power are considered at best "conjectures", or at worst "pseudoscience". Because they cannot be tested or falsified in any way, there is no way to determine whether they are true or false, and so they do not gain the status of "scientific theory".

Theories whose "predictive power" presupposes technologies that are not currently possible constitute something of a grey area. For example, certain aspects of string theory have been labeled as predictive, but only through the use of machines that have not yet been built and in some cases may never be possible. Whether or not this sort of theory can or should be considered truly predictive is a matter of scientific and philosophical debate.

This section does not cite its references or sources. Please help improve this article by introducing appropriate citations. (help, get involved!) This article has been tagged since March 2007.

One of the most famous cases of predictive power was the confirmation of Albert Einstein's theory of general relativity. The theory predicted that if photographs taken of the stars near the edge of the Sun during a solar eclipse were compared to photographs of the same stars when they were not near the Sun, a "bending" of their light by the Sun's gravitational field would be observed. The existing Newtonian theory of physics also -- for different reasons -- predicted some bending, but to a lesser degree. The prediction was made by Einstein in 1915 as a logical outcome of his theory, but it could not be tested until a solar eclipse on May 29, 1919, when observations were made by the astrophysicist Arthur Eddington which seemingly confirmed Einstein's predictions. The news was heralded throughout the world as a revolution in physics, as the Newtonian theory had been conclusively disproved.

The historical reality of the eclipse experiment is more complicated than the textbook account, and highlights some of the problems involved in retrospectively applying a notion such as predictive power. The results of the eclipse observations were far from clear – they were taken at two remote locations (Sobral in Brazil and the Atlantic island of Príncipe) and thus the telescopes used required accuracy to be sacrificed in favour of portability. In addition, there were mitigating factors such as the Earth's slight rotation during the eclipse, and the temperature differences between day (when the eclipse pictures were taken) and night (when the control pictures were taken), which caused optical anomalies. From the start, the experiments were far from clear-cut, and relied on a series of assumptions and human judgements.

Moreover, the data from the observations were not as conclusive as was professed. Two telescopes were used at Sobral; one produced 8 photographic plates which recorded a mean deviation from the norm of 1.98″ of arc (1 ″ = 1/3600^th of a degree), and the other 18 plates with a mean deviation of 0.86″; the two plates from a single telescope at Principe, though of a poor quality, suggested a mean of 1.62″. Einstein's theory suggested a deviation of approximately 1.75″, while Newton's suggested 0.8″. If all the data had been included, the results would have been inconclusive at best, but Eddington discounted the results obtained from the second Sobral telescope, claiming "systematic error", and gave extra weight to the results from Principe (which he had personally recorded), with little justification or supporting evidence. The Astronomer Royal, Sir Frank Dyson, and the president of the Royal Society, J. J. Thomson, sided with Eddington, and on November 6 declared the evidence was decisively in favour of Einstein's theory; much of the scientific community fell in line and agreed. Nevertheless, there were many scientists at the time who felt there were good reasons to doubt whether the prediction had been accurately fulfilled, or whether the results did not have an alternative interpretation. Subsequent eclipse observations in the 1920s and 1930s failed to provide confirmation, although many other different experiments have since provided much stronger (but less dramatic) proof of relativity.

The 1919 eclipse is one example used in science studies as a demonstration that scientific facts are in various ways constructed, or at least influenced, through a variety of assumptions, institutional forces, and interpersonal relations, and are rarely without expert dispute in their day. The philosopher and historian Thomas Kuhn has famously pointed out that "textbook" histories of science tell the story of the current theory as a linear set of triumphs, when in reality the historical record is much more complicated.

Other examples of predictive power of theories or models include Dmitri Mendeleev's use of his periodic table to predict previously undiscovered chemical elements and their properties (though largely correct, he misjudged the relative atomic masses of tellurium and iodine), and Charles Darwin's use of his knowledge of evolution by natural selection to predict that because there existed a plant (Angraecum) with a long spur in its flowers, a complementary animal with a 30 cm proboscis must also exist to feed on and pollinate it (twenty years after his death, a form of hawk moth was found which did just that).

The predictive power of a theory is closely related to applications.

General relativity does not only predict the bending of light as described above, but also predicts the proper time of satellites. This prediction is used to calculate positions via GPS.

If a theory has no predictive power, it cannot be used for applications.

• Accounts of the problems with the eclipse observation:
  • Harry Collins and Trevor Pinch (1993). The Golem: what everyone should know about science. Cambridge, England: Cambridge University Press. ISBN 0-521-47736-0.
  • John Waller (2003). Einstein's Luck: The Truth behind Some of the Greatest Scientific Discoveries. Oxford, England: Oxford University Press. ISBN 0-19-860719-9.