In the context of use case modeling the use case generalization refers to the relationship which can exist between two use cases and which shows that one use case (child) inherits the structure, behavior,  and relationships of another actor (parent).  The child use case is also referred to the more specialized use case while the parent is also referred to as the more abstract use case of the relationship.

For those of you familiar with object oriented concepts: use cases in UML are classes and the generalization is simply the inheritance relationship between two use cases by which one use case inherits all the properties and relationships of another use case.

You can use the generalization relationship when you find two or more use cases which have common behavior/logic.  In this instance, you can describe the common parts in a separate use case (the parent) which then is specialized into two or more specialized child use cases.

Example:

If you are creating a payment system which allows students of a training provider to pay for courses both on-line and by phone, there will many things in common between the two scenarios: specifying personal info, specifying payment info, etc.  However, there would also be differences between the two.  So, the best way to accomplish this is to create one use case (the parent) which contains the common behavior and then create two specialized child use cases which inherit from the parent and which contain the differences specific to registering on-line vs. by phone.