I see experimentation as a method for generating a novel action by more or less randomly modifying an action that is derived from an existing skill. Experimentation thus plays a similar role as the more or less random mixing of the two DNA patterns from the parents. For DNA, the second form of introducing random variations is through mutation. Again, a similar kind of process might happen within the individual to introduce a random variation into an action derived from a skill.
We assume that the action plan, i.e. the pattern underlying the action, is preserved, so that it can be used to modify a skill or instroduce a new skill if the experiment was successful.