How does a revolution in science come about? The answer: By trying to change as little as possible; by concentrating all efforts on the solution of a special and obviously still unsolved problem, and proceeding as conservatively as possible. For only where the novel is forced upon us by the problem itself, where it comes in a sense from outside and not from ourselves, does it later have the power to transform. And then, perhaps, it will bring very extensive changes in its train. [pp.151]
The object of this ‘as little as possible’ change is, of course, existing theory.
And then, in another essay in the same volume, ‘Changes of Thought Patterns in Science’, which briefly examines the history of the development of the special theory of relativity and quantum mechanics, Heisenberg, after describing how recalcitrant phenomena were only able to bring about a series of gradual, piecemeal, but ultimately extensive changes in the structure of physical theory, continues:
[N]ever in its history has there been a desire for any radical reconstruction of the edifice of physics. At the beginning, rather, we always encounter a very special, narrowly restricted problem, which can find no solution within the traditional framework. The revolution is brought about by researchers who are genuinely trying to solve this special problem but who otherwise wish to change as little as possible in the previously existing science. It is precisely the wish to change things as little as possible which demonstrates that the introduction of novelty is a matter of being compelled by the facts; that the change of thought pattern is enforced by the phenomena, by nature itself, and not by human authorities of any kind. [pp. 163-164]
There is plenty to quibble about here in Heisenberg’s description of scientific theory change as being driven by ‘facts’ and ‘nature’ and ‘not by human authorities of any kind.’ (Heisenberg does acknowledge the psychological resistance of physicists to having established research programs change, and notes that new theories are only taken on board when they are more ‘fruitful’ and can lead to greater ‘success’. Theory change appears a little more expedient here.)
Be that as it may, my primary interest in excerpting these two passages here was rather more personal. My doctoral work was in belief revision–the formal modeling of the process of rational belief change–and a fundamental methodological principle guiding the design of logical operators that accomplished the various species of such change was the principle of minimal change. That is, a rational agent, when accepting–or ‘forced’ to accept–a new piece of information that conflicts with its older beliefs, will strive to give up as few of its older beliefs as possible. This principle was said–perhaps by Isaac Levi, though my memory fails me now–to find its grounding in the history of science, but I do not think I was ever directed to Heisenberg’s formulations of it. The histories of special relativity and quantum mechanics, often cited by Kuhnians as examples of radical reconstruction in science, offer, I think, good empirical support for Heisenberg’s understanding of the process of theory change in science.