A learner or change target can be highly motivated to learn something, yet have no role models nor initial feeling for where the answer or solution might lie. The learner then searches or scans by reading, traveling, talking to people, hiring consultants, entering therapy, going back to school, etc. to expose him or herself to a variety of new information that might reveal a solution to the problem. Alternatively, when the learner finally feels psychologically safe, he or she may experience spontaneously an insight that spells out the solution. Change agents such as process consultants or non-directive therapists count on such insights because of the assumption that the best and most stable solution will be one that the learner has invented for him or herself.
Once some cognitive redefinition has taken place, the new mental categories are tested with new behavior which leads to a period of trial and error and either reinforces the new categories or starts a new cycle of disconfirmation and search. Note that in the process of search, if role models are readily available, they will most likely be used. Identification is thus an efficient and fast process, but it may lead to solutions that do not stick because they do not fit the learner's total personality. If one wants to avoid that, one must create learning environments that do not display role models, thereby forcing the learner to scan and invent his or her own solutions.
It is this dynamic, to rely on identification with a role model, that explains why so many consultation processes go awry. The consultant, by design or unwittingly, becomes a role model and generates solutions and cognitive categories that do not really fit into the culture of the client organization and will therefore only be adopted temporarily. A similar result occurs when organizations attempt to check on their own performance by "benchmarking," i.e. comparing themselves to a reference group of organizations and attempting to identify "best practices." The speed and simplicity of that process is offset by two dangers. First, it may be that none of the organizations in the reference set have scanned for a good solution so the whole set continues to operate sub- optimally, or, second, that the identified best practice works only in certain kinds of organizational cultures and will fail in the particular organization that is trying to improve itself. In other words, learners can attempt to learn things that will not survive because they do not fit the personality or culture of the learning system. For change to remain more stable it must be "refrozen."