Science 240 Fall 1998

The Nature Of Scientific Thought

The Logic of the Standard Hypothetico-Deductive Model of Hypothesis Testing

1. Discovery vs. Justification
   • Scientific theories must be testable
2. Eddington's test of Einsteins' Theory of General Relativity

   General strategy: Test a theory about "unobservables" by looking for its "observable" consequences. I.e. A theory describes an underlying process or pattern (the causes)which explains observable phenomena (the effects). A true theory has to correctly describe the observable effects.

   1. Identify the Logically relevant Test Components
      1. The theory or hypothesis under test.
      2. Procedures used in obtaining data, preparation of the samples, etc.
      3. The data or observations predicted by the theory (i.e. the test prediction)
      4. Other theories or assumptions used in obtaining the prediction or in obtaining the data (called background assumptions)
      5. The actual data obtained
   2. Striking or impressive feature about Eddington's test?
      1. Why regard the data as a striking confirmation of Einstein's theory?
      2. The meaning of confirmation
3. Guidelines in Designing a Good Test
   1. General strategy: Design the test so that a true prediction is strongly linked to a true hypothesis, and a false prediction is strongly linked to a false hypothesis.
   2. Guidelines
      1. The test prediction must be a logical implication of the theory under test given the test-set-up. (when combined with the background assumptions) I.e.- The theory requires the test prediction, and is actually used in making the prediction.

         * This condition is needed to hold the theory responsible for a false prediction.

      2. The prediction should be "risky", and not easy to get right by some means other than the theory in question.
         1. The prediction should be precise
         2. The more unusual the prediction the better
         3. There should be no other available theories which make the same test prediction. I.e. no other available explanations of a correct prediction.

            - Requires proper controls of other variables that could produce the data

            * These conditions are needed to give the theory under test full credit for a correct prediction.

4. Drawing Conclusions about the Theory under Test
   1. The prediction is true.
      1. It's a good test. Conclude that the theory gets a large boost in probability of being true.
      2. It's not a good test. No boost in confirmation. Insignificant data.
   2. The prediction is false.
      1. The theory required the prediction to be true. Conclude that the theory is false.
      2. The theory did not require the test prediction to be true. The theory has not been refuted.
5. Examples.
   1. The Wave Theory of Light: 1818

      After a century in the shadow of Newtonian ideas, according to which light is composed of small particles, wave models of light were revived around 1800, first by an Englishman, Thomas Young, then by a Frenchman, Augustin Fresnel. Fresnel's model was submitted for a prize offered by the French Academy of Science, One of the judges for the Academy, S.D. Poisson, deduced that, according to Fresnel's models, the shadow of a small circular disk produced by a narrow beam of light should exhibit a bright spot right in the center of the shadow. Poisson and the other judges are reputed to have thought that this refuted Fresnel's hypotheses because they had never heard of there being such a phenomenon and regarded it as highly unlikely to exist. No known particle models predicted such a spot. When the experiment was carried out in carefully controlled circumstances, however, there was the spot, just as required by Fresnel's model. Fresnel received the prize in 1818.

   2. Pyramid Power

      The following case has been cited in support of the hypothesis that pyramids have special powers. A young woman who was having difficulty with her complexion was told to keep a pitcher of water under a pyramid and then wash her face in that water, with only the mildest soap, once in the morning and once in the evening. She was also told to put nothing else on her face, no creams or medications of any kind, and no makeup. Although she had been in the habit of using quantities of makeup, she agreed to the experiment. Within two weeks there was a clearly noticeable improvement in her complexion.