The model should include a constant effect, such as ”all subjects”. Tis is typically implicit in SPM. If your model does not include the constant, the factor that you are interested in (such as patients > controls) will model both the overall effect (such as default mode connectivity) and the difference between patients and controls, which is probably not what you are looking to do.
When you define your contrast at the second level, chose ”all subjects” and the factors you are interested in, and then select the contrast from the drop-down list that you wish to evaluate.