Here, we argue that action learning (AL) has been evolving into different variations, whose respective advocates appear to concentrate on one of the several components inherent in Revans' formulation of AL as L = P + Q. They do this--sometimes inappropriately--to the virtual or relative exclusion of other aspects, and this has consequences for the outcomes and impact of the AL process. In an attempt to delimit the boundaries between various versions and indeed to identify what Johnson [2010. A framework for the ethical practice of action learning. "Action Learning: Research and Practice" 7, no. 3: 267-283] called "inauthentic" AL, we have been developing our ideas for a scanning device or framework. We refer briefly to some of the theoretical underpinnings of this framework. We then introduce a fresh taxonomy to explain and illustrate features of five principal variations of emphasis in AL that we have identified. The aim of this framework is to help stakeholders to work towards selecting and co-creating the most appropriate variation of "authentic" AL to suit their unique set of circumstances at any given time. We outline the likely outcomes of each respective variation if taken to extremes and conjecture about their implications. This taxonomy should also help one to reduce the mystique and confusion that often surround AL while acknowledging its complexity. We suggest that by taking advantage of insights provided by this framework, purchasers and potential AL set members in particular are more likely to participate in learning conversations that lead to more informed decisions and actions to address or adjust their respective interests and needs. In conclusion, we identify some areas for further research and development. (Contains 2 figure, 5 tables and 5 notes.)