If X is a sufficiently nice connected topological space, then H^1(X,Z/n) should correspond to morphisms \pi_1(X) -> Z/n. This should just be connected Z/n-covering spaces, right? . On the other hand, we can interpret H^1(X,Z/n) as being given by principal homogeneous spaces. Do these also have to be connected if X is? What is stopping me from say taking a disjoint union of two connected Z/5-covering spaces and letting Z/10 \cong Z/2 x Z/5