A few notes on Hypotheses and Predictions

By Peter Grant, Princeton University, USA
(adapted by Dieter Ebert)

Definitions

In the formal language of standard scientific methodology an hypothesis is an explanation for a phenomenon. It is based on one or more assumptions. Often the assumptions are not stated but simply implied.

A prediction is a logical consequence of the hypothesis, i.e. an expectation derived from the hypothesis.

Typically an hypothesis is tested by determining if a prediction is correct or not. We say if our hypothesis A is correct we expect to observe X. If we then observe X the hypothesis is supported. If instead we observe Y the hypothesis is not supported, it is rejected. Alternatively the hypothesis may be discarded, or modified, if the assumptions are not met.

Examples

1. A single hypothesis

Each year male marine iguanas change color at the beginning of the breeding season. Some develop red coloration on their (black) bodies while others develop green patches. One hypothesis (A) for the variation in color is that it is caused by diet variation; red-colored iguanas eat different algae from green-colored iguanas. The hypothesis predicts a close association between algal type and iguana color during the development of breeding conditions. The hypothesis is rejected if there is no such association, and another hypothesis will have to be constructed. On the other hand if the predicted association is found, a second prediction can be tested with controlled experiments: red iguanas raised on green iguana diet should develop green color and green iguanas raised on red iguana diet should develop red color. Note the second test is stronger than the first; the second prediction tests the causality of the hypothesis whereas the first prediction tests the necessary condition of correlation between color and the environment. The second test assumes that colours, once developed, can change again.

2. Two contrasting hypotheses

In the above example the investigation tests the possible role of a single factor in explaining observations. Often a phenomenon can be explained by more than one factor, and then the task for the investigator is to choose between alternative explanations. A genetic hypothesis (B) could be offered to explain the observed variation in iguana color: red-colored iguanas have a different genotype from the green-colored ones. The challenge (and the fun!) is to set up experiments, or a program of planned observations, to distinguish unequivocally between the alternative hypotheses by rejecting one of them. How could hypothesis A, or B, be rejected? (Note: Having two hypotheses does not exclude third or fourth options).