The way I'm used to programming behavior for objects is: subclass "GameObject", and override the updateLogic function.
In updateLogic, I had access to certain information by default (mostly provided by the physics engine), like a list of objects that this object is colliding with, the current velocity vector, position vector, rotation matrix/quat, etc - along with other functions that allowed me to modify the object state.
Now, I looked at the examples that come with JME, and I'm pretty sure that I can create that setup on my own (a game object that wraps/extends a scene node, and a main loop that would run "updateLogic" on each and every game object), but, since this is a game engine, rather than just a rendering engine, I would hope that doing so would be unnecessary, and that JME provides a standardized mechanism for executing object specific logic efficiently.
Does it?
I mean, I didn't see one example where things were split through multiple files, each defining a game object (or a JME equivalent) with specific logic functions.