The Java API documentation states that, for an object x, the general intent of the clone method is for it to satisfy the following three properties:

• x.clone() != x (the cloned object is a different object instance)
• x.clone().getClass() == x.getClass() (the cloned object is the same type as the source object)
• x.clone().equals(x) (the cloned object has the same 'contents' as the source object)

For the cloned object to be of the same type as the source object, non-final classes must call super.clone and that call must eventually reach Object.clone, which creates an instance of the right type. If it were to create a new object using a constructor, a subclass that does not implement the clone method returns an object of the wrong type. In addition, all of the class's supertypes that also override clone must call super.clone. Otherwise, it never reaches Object.clone and creates an object of the incorrect type.

However, as Object.clone only does a shallow copy of the fields of an object, any Cloneable objects that have a "deep structure" (for example, objects that use an array or Collection) must take the clone that results from the call to super.clone and assign explicitly created copies of the structure to the clone's fields. This means that the cloned instance does not share its internal state with the source object. If it did share its internal state, any changes made in the cloned object would also affect the internal state of the source object, probably causing unintended behavior.

One added complication is that clone cannot modify values in final fields, which would be already set by the call to super.clone. Some fields must be made non-final to correctly implement the clone method.