Dear Aisling,

I also ran experiments with subjects having different number of slices, it is for sure possible. You just need to preprocess each of them separately, with inputting the correct number of slices in SPM. After all the preprocessing steps, the brains will be normalized to MNI, so there will be no subsequent issue in CONN. Also as an alternative, if your TR < 2s, you can just skip slice timing correction, it does not make much of a difference in practice (even though in theory it still enhances the SNR).

Hope this helps,
Stephen