This question is motivated by the statement in "Differential Topology", by Guillemin and Pollack, page 133:
Patterns admissible on the torus, like one saddle plus one source, are prohibited on the sphere.
Here, patterns refers to the behaviour of a vector field about its zeros. Of course, using the Euler Characteristic, we can rule out the existence of such a vector field on the sphere, however it is not clear to me whether such a vector field exists on the torus.