By referring to the preceding spinach example, the null hypothesis would be "Eating spinach has no effect on long-term memory performance".

The null hypothesis is just that—it nullifies what we're trying to prove by running our experiment. It does so by asserting that a statistical metric (to be explained later) is zero.

The alternative hypothesis is what we hope to support. It is the opposite of the null hypothesis and we assume it to be true until the data provides sufficient evidence that indicates otherwise. Thus, our alternative hypothesis, in this case, is "Eating spinach results in an improvement in long-term memory".

Symbolically, the null hypothesis is referred to as H0 and the alternative hypothesis is referred to as H1. You may wish to restate the preceding null and alternative hypotheses as something more concrete and measurable for our study. For example, we could recast H0 as follows:

"The mean memory score for a sample of 1,000 subjects who ate 40 grams of spinach daily for a period of 90 days would not differ from the control group of 1,000 subjects who consume no spinach within the same time period."

In conducting our experiment/study, we focus on trying to prove or disprove the null hypothesis. This is because we can calculate the probability that our results are due to chance. However, there is no easy way to calculate the probability of the alternative hypothesis since any improvement in long-term memory could be due to factors other than just eating spinach.

We test out the null hypothesis by assuming that it is true and calculate the probability of the results we gather arising by chance alone. We set a threshold level—alpha (α)—for which we can reject the null hypothesis if the calculated probability is smaller or accept it if it is greater. Rejecting the null hypothesis is tantamount to accepting the alternative hypothesis and vice versa.