Recent enthusiasms have emphasised particularly ideas of “learning organisations”, calling attention both to the ways organisations learn and to the possibilities for improving organisational intelligence through increasing the rate and precision of learning. These enthusiasms have often encouraged the too easy assumption that learning processes necessarily lead to organisational improvement. The usefulness of learning as an instrument of organisational intelligence has to be demonstrated, not assumed. Adaptive processes such as learning are not guaranteed to reach or sustain a global maximum on an outcome surface. In order for proposals for “learning organisations” to be more than vague wishes for improvement, learning must be defined in terms of some specific process and the conditions under which that process does or does not lead to improvement must be established. This essay is built around six simple propositions, a kind of catechism for consideration of learning as an instrument of intelligence.