Good experiments have the potential to falsify hypotheses

Once predictions are made from a hypothesis, experiments can be designed to test those predictions. The most informative experiments are those that have the ability to show that the prediction is wrong. If the prediction is wrong, the hypothesis must be questioned, modified, or rejected.

There are two general types of experiments, both of which compare data from different groups or samples:

  1. Controlled experiments manipulate one factor of interest while holding other variables constant as a means of testing the influence of the manipulated variable.

  2. Comparative experiments compare data gathered from different populations that differ in multiple, unknown ways.

Rachael Bay and her colleagues conducted both types of experiments, to reveal genetic differences between the coral populations as well as their abilities to acclimate to different temperature environments.

In a controlled experiment, we predict on the basis of our hypothesis that some critical factor, or variable, has an effect on the phenomenon we are investigating. We devise some method to manipulate only that variable in an “experimental” group and compare the resulting data with data from an unmanipulated “control” group. If the predicted difference occurs, we then apply statistical tests to ascertain the probability that the manipulation caused the difference (as opposed to the difference being the result of random chance). Investigating Life: Corals in Hot Water describes a controlled experiment examining the effect of pool of origin on the response of corals to heat stress done by Rachael Bay and other members of the Palumbi laboratory. A good controlled experiment is not easy to design because biological variables are so interrelated that it is sometimes difficult to alter just one. In this experiment, the controlled variable was whether the experimental corals came from the cool-pool or the warm-pool environment. They were then subjected to the same thermal stresses and their responses measured as extent of bleaching. The control groups were corals not exposed to the thermal stress protocol.

A comparative experiment starts with the prediction that there will be a difference between samples or groups based on the hypothesis. Unlike in a controlled experiment, we cannot control all the variables in a comparative experiment; often we cannot even identify all the variables involved. We are simply gathering and comparing data from different sample groups. Rachael and her colleagues did a comparative experiment in addition to the controlled experiment shown in Investigating Life: Corals in Hot Water. They analyzed and compared the genetic makeup of corals taken from the warm and cool pools.