The process of constructing of a theory from several individual theories can be regarded as a form of learning, because the system<<s internal representation gets reformulated and this leads to improvements of performance. We notice that the reformulations are rather simple. They involve inclusion (or non inclusion) of rules in the integrated theory and modification of coefficients affecting their selection. More complex forms of theory reformulation could also involve modifications of individual rules. Despite this simplicity the method leads to noticeable improvements in performance. More work can, of course, be done in this area. In the following we shall discuss some possible extensions of this work.