Null Hypothesis testing is often credited to Sir Ronald Fischer. In 1919 Fisher started work at a small agricultural station. Here he analyzed extensive collections of agricultural data. This investigation resulted in a series of reports in which he pioneered the principles of null hypothesis testing and analysis of variance techniques (among many other statistical discoveries). Thus, the null hypothesis test was created to compare different agricultural techniques.
Null hypothesis testing has always been controversial. Many statisticians have pointed out that rejecting the null hypothesis says nothing about the null hypothesis being true. Under traditional null hypothesis testing, the null is rejected when the data collected seems extremely unlikely. However, researchers are often interested in whether a hypothesis is true. Null hypothesis testing can only provide evidence for a hypothesis because the data says that the opposite hypothesis is very unlikely.
Consider if I won the lottery. I could be accused of cheating by bribing lottery officials. At a trial, the prosecutor might point out that my winning of the lottery without cheating is extremely unlikely. My being innocent is therefore comparably unlikely (rejects null hypothesis). This reasoning is faulty because there is an extremely low probability that anyone would win the lottery. I may have or may have not cheated to win the lottery, but because of the low likelihood of the event the null hypothesis testing method would tell us nothing about the truth of the statement.
Many people have noted that estimates of the magnitude of the effect and/or confidence intervals would provide much greater information than a null hypothesis test.
Jacob Cohen has written extensively on null hypothesis testing and effect size. Below is a reference for one of his most interesting articles on this controversy:
Cohen, J. (1994). The earth is round (p<.05 american psychologist>