How to Conduct a Scientific Test

Ken Kostka
American Swallow Conservancy
Pittsburgh, PA

You do not need a degree in a scientific discipline to design a scientific experiment. Scientific tests do not always require elaborate planning and complex mathematical skills. In fact, anyone with average intelligence and a little common sense can conduct a valid scientific experiment as long as the following scientific principles are adhered to:
Hypothesis formulation: What is the question for testing? An hypothesis leads to a series of predictions. Examples of a hypothesis: 1. Martins prefer round holes over starling resistant holes. 2. Treating nests with Sevin (carbaryl) increases breeding success in Purple Martins. 3. Martins breed more successfully in houses than in gourds. 4. Martins are more likely to nest at a site where the dawnsong is being broadcast.

Designation of the variable : What factor will be manipulated? Only one thing should be manipulated, so that any changes can be attributed to that one factor. For example, suppose that you have two hypotheses you want to test 1) Broadcasting dawnsong increases the chance of attracting breeding martins and 2) Deploying decoys increases the chance of attracting breeding martins. If the experimenter were to both play dawnsong and use decoys at the same group of test sites, and breeding martins were attracted more frequently to this group of sites, it would be impossible to say whether the decoys or the dawnsong was the variable that influenced the attraction of martins. Always test variables independently.

Control groups: Always have control groups. Controls are an unmanipulated set of test subjects, used so the investigator can see what would happen without the experimental condition. For example, when Morton hypothesized that broadcasting the dawnsong would increase the chances of attracting breeding martins, he tested this hypothesis by playing the dawnsong at some sites, and by not playing it at an equal number of other sites with similar martin housing and site characteristics. The only difference between the two groups of sites was that the dawnsong was being played at one group of sites. If Morton had not used control groups, and martins were attracted to the sites where the dawnsong was being played, one could argue that the martins would have nested at those sites anyway (even if the dawnsong had not been played).

Sample size: While even a sample size of one is valid, the larger the sample size, the more convincing the results. Some hypotheses are very difficult to test because it is difficult to get a large and convincing sample size. For example, if you are attempting to test for martins’ preferences of martin house characteristics at uncolonized (unestablished) sites in areas where martins are scarce and difficult to attract, the attraction of each newly attracted pair of breeding martins to any new site will be one sample. To get a sample size of ten, you would need to attract ten pairs of breeding purple martins to new sites. (either 1 pair to 10 new sites or 10 pairs to one new site or some combination)

Other factors to consider:
– The next time you are told or assume something about martins, ask if it has been scientifically tested. There may be a lot of anecdotal evidence, which is highly suggestive, but no conclusive scientific evidence. Some of the most basic assumptions have never been tested scientifically.
– Consider who did the testing and look at the data. Experiments can be manipulated. The experimenter may have “bias.” For example, if a person conducts a test to determine whether martins prefer houses or gourds for nesting, but the researcher manufactures martin houses, he or she may intentionally or unintentionally manipulate the experiment to obtain the desired “results.”  The same could be said if the experimenter manufactured gourd housing. To the best of my knowledge, no scientific experiment has ever been conducted to test martins’ preference for houses or gourds in any geographic area
– Even a properly conducted scientific test is not necessarily definitive. In fact, one test of whether results are valid is whether they can be repeated.