p≤0.10, p≤0.05, p≤0.01

Researchers use inferential statistics to calculate the probability of obtaining their results if the null hypothesis was true. If this probability (chance) is really low, i.e it is really unlikely that you would have got these results, then it suggests that the null is untrue and therefore should be rejected. But how unlikely does the probability need to be? In psychology, like many other sciences, we use the 0.05 level of significance, which means that the probability of obtaining our results if the null were true must be equal to or less than 1 in 20 (5%) for us to reject the null hypothesis. If this is not the case and the p value is greater than 1 in 20 (p>0.05) then we retain the null hypothesis. This is because the size of the p value suggests that obtaining our results is quite likely if the null was true, therefore we must retain it. We are not saying that the experimental hypothesis is incorrect and the null is true just that we currently do not have enough evidence to reject the null hypothesis, i..e to say these two variables are unrelated. 

If our results are significant at the 5% level, (p<0.05) we may decide to check for significance at the 1% level. This would mean that the portability of obtaining our results if the null were true would be less than 1 in 100 (1%). This is exactly what was found in the study by Goldfarb et al. (1943) when they tested the maternal deprivation hypothesis. Their results were significant at p<0.01. The probability of there being no difference between the average scores for cognitive development for children adopted earlier rather than later was 1 in 100. This is so unlikely that the null hypothesis is rejected and the experimental/alternative hypothesis accepted.