D Hypotheses Are Not Predictions

Although hypotheses and predictions are often linked, they are different concepts, and failing to distinguish between them creates needless confusion. A hypothesis is a putative explanation for actual observations.

A prediction is a forecast of how some future event will play out. We test predictions directly with experiments that show whether they are true or false. We test hypotheses indirectly by testing their predictions. By making predictions, the hypothesis reveals exactly what it does and does not mean, establishes a ra­tionale for the conduct of investigations, and specifies the relationships among relevant variables. You might hypothesize that your new puppy, Snowball, jumps around hysterically when you return home because he has never learned to act in a more civilized manner. One prediction of your hypothesis would be that Snowball would behave himself better if he were properly trained. The hypo­thesis itself is not directly testable (you cannot measure the state of learning in the relevant parts of Snowball's nervous system); you can test it indirectly, how­ever, by sending Snowball to Puppy School to be trained.

Because predictions are logical deductions from hypotheses, if a hypothesis makes a prediction, then the prediction must be true. It does not work the other way, though: true predictions do not allow you to conclude that your hypo­thesis must be true. If Snowball's welcoming activity decreased after training it might mean that he simply got used to your returning home; it was no longer an amazing event, and the training itself did nothing. Or the problem might not have been his lack of self-control. You, his owner, might have been reinforcing his over-the-top greeting by getting down and rolling around on the floor with him. Maybe the training just got you to calm down. Your prediction (he'll behave better after training) would be correct, while your hypothesis to explain it (the dog is the problem) could still be wrong.

People sometimes say they “hypothesize” about what the outcome of a meas­urement will be. What they mean is that, while they haven't done the measure­ment and don't know how it will turn out, they do have a guess that they call a hypothesis. However, their guess does not explain anything; it is not a hypothesis and merely indicates that there is something they want to measure. For example, you might wonder if your tap water is acidic and guess that it might be. There is no hypothesis involved; you get a pH meter, take a sample of the water, and de­termine its acidity. You get an answer: the pH is or is not more acidic than 7.0. Predictions like this are not related to hypotheses. And it is not a question of being able to do the test right away (maybe you don't happen to own a pH meter and can't test the water); what is fundamentally a prediction does not turn into a hypothesis merely because the measurement can't be carried out.

The difference between hypothesis and prediction depends on why you're doing a measurement, not the measurement itself. Identical measurements might test hypotheses or predictions. One day you might think, “Dang, my tap water sure tastes funny; I'll bet it's because of that acid rain I've been reading about.” In this case your hypothesis is explanatory—the tap water's taste is dif­ferent because acid rain has tainted it—and extremely complex.

Whatever the final answer, the acid rain hypothesis would predict at least that the pH of your tap water would be acidic and you could test this prediction with the pH meter. (Even “pure” rain has a slightly acidic pH; nevertheless acid rain, at approximately pH 4.0¹⁴ is more acidic still.) If the pH of your water is not acidic, then its taste probably wasn't caused by acid rain, and you should entertain other hypotheses: the chemical plant upstream, etc. Likewise, if the pH of your tap water was acidic, that would be consistent with your hypothesis but not prove it was correct. You might be encouraged to test other predictions (e.g., that the source of your water—lake, reservoir, water tower on top of your building—is acidic and that it is acidic because of the rain). The difference between prediction and hypothesis has nothing to do with what measurement is made, but with why it is made and what conclusions it allows.

A hypothesis is an explanation for a phenomenon, and I've suggested that it is often expressed in terms of properties or causes that we can't directly observe or measure. The reason is that, if we could directly measure or observe them, then that's what we'd do—no reason to beat around the bush with hypotheses. But what do the concepts “direct” and “indirect” mean? I've been using them in a log­ical context—we test hypotheses indirectly via their predictions, and predictions are statements about the world that we can test directly by making measurements. There is also a practical sense in which scientists refer to these terms.

2.