The theory advanced to explain why some phenomenon occurs cannot be tested. We test a theory by means of its predictions. For example, a theory may predict that light bends under certain conditions, or that a liquid will change colour if sprayed with acid, or that a psychotic person will respond badly to particular stimuli. If the predicted event fails to occur, then this is evidence against the theory. A thoery cannot be tested when it makes no predictions. It is also untestable when it predicts events which would occur whether or not the theory were true.
- Aircraft in the mid-Atlantic disappear because of the effect of the Bermuda Triangle, a force so subtle it cannot be measured on any instrument. (The force of the Bermuda Triangle has no effect other than the occasional downing of aircraft. The only possible prediction is that more aircraft will be lost. But this is likely to happen whether or not the theory is true.)
- I won the lottery because my psychic aura made me win. (The way to test this theory to try it again. But the person responds that her aura worked for that one case only. There is thus no way to determine whether the win was the result of an aura of of luck.)
- NyQuil makes you go to sleep because it has a dormative formula. (When pressed, the manufacturers define a "dormative formula" as "something which makes you sleep". To test this theory, we would find something else which contains the domative formular and see if makes you go to sleep. But how do we find something else which contains the dormative formula? We look for things which make you go to sleep. But we could predict that things which make you sleep will make you sleep, no matter what the theory says. The theory is empty.)