A hypothesis (plural: hypotheses) is in its simplest form nothing more than an idea about how the world works. For example, “the moon is made of green cheese” is a valid hypothesis. But there are several characteristics which separate useful scientific hypotheses from those which are impractical.
First and foremost, a hypothesis must be testable. We must (at least in principle) be able to design an experiment which will allow us to determine whether the hypothesis is false. Keep in mind that we can never prove any hypothesis is completely true because we can always
English chemist Robert Boyle (1627-1691) was one of the first scientists to explicitly adopt a program of hypothesis testing. imagine new circumstances in which it has not been tested, or other possible explanations for the results we have obtained. It is much easier to show, with a high degree of confidence, that a hypothesis is false. If it is not consistent with the results of a well designed and executed experiment, we are forced to accept that the hypothesis is false. If a hypothesis is not falsifiable, it is outside the realm of science. Note that the “green-cheese hypothesis” meets this test.
A hypothesis should also be plausible. That is, the hypothesis, should be consistent with what we already know about the subject being investigated, and its parts should be logically and mathematically sound. We often celebrate the creative spark by which new hypotheses come to light. But typically that moment of inspiration follows a great deal of perspiration racked up in a thorough review of previous research in the subject. A hypothesis may in the end be a guess, but it should be the best guess possible. Given what we know about astronomy (and cheese production) the”green-cheese hypothesis” is not plausible, and not worth investing much of our time and resources in testing.
What are predictions?
The predictions of a hypothesis set out what we expect to see if the hypothesis is true. (This is where we use deductive “If-Then” logic.) Experiments are designed to test specific predictions of the hypothesis. The “green-cheese hypothesis” predicts that material collected from the moon would contain milk proteins and fungi. These predictions could be tested by bringing material back from the moon, and testing its chemical structure. The hypothesis also makes predictions about the wavelengths of light reflected from the moon, a field called spectroscopy. (These predictions have actually been tested, believe it or not. Needless to say the hypothesis was not supported!) Three main factors make a prediction useful in testing a hypothesis:
The prediction should be specific to the hypothesis (i.e. no other hypotheses make the same prediction). If several hypotheses predict the same outcome of an experiment, we will need to do further experiments to distinguish between them.
The prediction should provide results which are unambiguous.
It should be practical and economically feasible to run the experiment.
A prediction is really nothing more than a simpler hypothesis — practical to test — derived from a larger hypothesis. Note that a prediction does not have be about the future, but it does have to apply to a situation we have not looked at yet. We are free to use the results of previous experiments to develop a new hypothesis, but we can’t then test our predictions against the results of those old experiments — to do so would be arguing in circles.
Are theories different from hypotheses?
A “theory” has no formal definition in science (Style Manual Committee, CBE 1994). Hypotheses which have considerable support from experiments, and which are useful in explaining a fairly wide range of phenomena, are “upgraded” to theories, for example Darwin’s Theory of Evolution by Natural Selection, or the Theory of Plate Tectonics (which explains the movement of continents). So a theory is simply a well-tested and widely useful hypothesis, and there’s no strict rules defining when a hypothesis becomes a theory.
Theories which are extremely well supported by experiments, particularly those which can be expressed as simple mathematical equations, are often called laws, e.g. Newton’s Law of Gravitation, Kepler’s Laws of Motion, or Mendel’s Laws of Genetics. Again, no one has yet laid out a strict set of rules for defining a natural law.
A final term that scientists use to describe their ideas is a model. This dates back to the time when physical models were one of the few tools researchers had in investigating phenomena which were too big or too small to manipulate directly. Physical models are still used in science. Francis Crick and James Watson used a scale model of a DNA molecule to help them deduce its structure (Giere 1997). But scientists also use mathematical models to help them understand how different factors will interact. The development of computers has vastly increased the scope of mathematical models, and made them accessible even to non-mathematicians.
Why are hypotheses important?
Philosopher of science Karl Popper likened a hypothesis to a searchlight, which the researcher shines on the relevant portion of nature (Davies 1973). It tells us which experiments are the important ones to perform, and which observations the important ones to make, out of an infinite number of possibilities. Without hypotheses scientists would be reduced to bean counters, and science to a collection of facts without organization or purpose. A hypothesis is the cornerstone used in building an elegant, structured body of knowledge from the apparent chaos of nature. (So they’re pretty important, eh!)