I saw some control diagrams for emotions on this website http://www.emotionalcompetency.com/sadness.htm and thought it would be cool to model it with a state space formalism. let's take x as a vector x = [anger, sadness, joy, etc...] where anger sadness and joy are quantities probabilities that one is angry, sad, joyful, etc.... the equation d x/dt = A x where a is the probability of transitioning between emotions. x would be normalizable to one. Since A is a probability transition matrix, it's unitary, so the dominant eigenvector of A would have eigenvalue of 1. This would an emotional eigenstate that doesn't change in time. All other eigenstates would oscillate or decay for negative or complex eigenvalues. A "psychon" would be a quantization of the amplitude in x. Granted feeling are non linear and this is a first approximation, but it would be cool to see where this goes.