Discovering Statistics Using IBM SPSS Statistics: North American Edition

by Andy Field

if you want to test for normality before a paired-samples t-test then you should compute the differences between scores, and then check if this new variable is normally distributed as a proxy for the sampling distribution