I am trying to understand the rules defined in the UML2 metamodel for associations between members and Usecases in Usecase diagrams. I understand that actors and usecases can participate in generalizations (because they are derived from a classifier that adds the ability to generalize), and I understand that usecase can have an extension / include relationship with other usecases (as this is explicitly specified in the specification for Usecase charts). Here is part of the metamodel diagram for usecases:

As you can see, there are no connections between actors and customs in this diagram. This can only mean that associations are allowed due to the general supertype of the actor and usecase, possibly a BehavioredClassifier or Classifier. So my question is:
- Which of the meta-classes or elements in the hierarchy of the metamodel is responsible for the associations between subjects and customs?
Another thing that is confusing: there is a limitation that states:
An actor may have associations for use only in cases, components, and classes. In addition, these associations must be binary.
But there is no equivalent restriction defined for usecases (except for the binary thing). Does this mean that associations always belong to actors, and in cases where they are not familiar with their associations with respect to actors? The specification can be found here , thanks.
ceran source share