|Eyes wearing |
Philosophy of science
An experiment is an integral part of the whole scientific method, whereby theories are tested by the experiment and then refined to explain new or surprising results. Sitting in an armchair and pondering can produce interesting ideas, but can't reflect reality until they are tried in reality. A good set of experiments should aim to demolish a theory by producing data that it cannot explain, while a good theory should be able to resist such things by being aptly able to explain the data.
The vast majority of experiments performed by scientists throughout the world are fairly straightforward and the theories behind them are well understood; a routine NMR experiment may throw up some new chemicals or new reactions, but is unlikely to shake the core theory of magnetic resonance. Similarly, experiments done to test evolutionary theories have so far failed to put a dent in the concept of evolution or the mechanism of natural selection, but they do sometimes change the niggling details about what genes do which specific task.
However, there are occasional surprises that cause scientists to rethink their theories almost from scratch. Experiments can also show that people's intuition can be wrong (or at least less right).
Controlled experiments are experiments in which as many factors as possible are kept unchanged. This is for attempting to extract the effects of that factor and to eliminate the effects of the other factors. Doing so helps avoid lengthy and often fruitless arguments about which factors were really involved in producing some result. Experimental subjects where some factor is absent are often called "controls", and they are compared to the subjects where the factor is present.
A classic controlled experiment was performed by Francesco Redi about 350 years ago on the spontaneous generation of flies. He was describing flies coming from rotting meat when he noticed that the meat attracts similar flies. Were the flies reproducing at that meat? He tried keeping flies away from some meat. It still rotted, but no flies appeared. But could it still attract flies? Could it receive some influence from the air? He put some rotting meat in some containers that he covered with gauze, and he discovered that the meat still attracted flies, though the flies could not get to it. But the meat did not make flies. In summary:
|Container state||Accessible to the air||Accessible to flies||Flies attracted||Flies produced|
The first experiment is the control, because it has no accessibility restrictions. The second one is unable to distinguish between these hypotheses: (1) keeping flies away prevents fly production and (2) keeping air away prevents fly production, or else keeps the meat from attracting flies. That may be why Redi did his third experiment, which allowed air to reach the meat but not flies. It distinguished between (1) and (2), thus establishing that it is flies that make rotting meat produce flies, and not anything in the air. In present-day jargon, access to air would be a "confounding factor".
- Randomized controlled trial
- For an example of an experiment that is so badly designed that it self-invalidates, Richard Hoagland's Accutron stunt