SPSS excluding whole cases when it shouldn't


I am doing a within-subjects analysis using the GLM Repeated Measures in SPSS. My analysis is a 3x9 and I have 21 participants. However, I noticed that the degrees of freedom for my F statistics were not the expected values. One of my colleagues told me that this is because SPSS just throws out the whole case (subject) even if only one data point is missing for that subject.

How do you stop SPSS from doing this?