JG,
I'd echo some of alf's responses to your post. First, CRM is not an 'object' or defined property of the operational environment. It's a broad spread of themes drawn from the social analysis of the conduct of work. The only attempt to define it (Lauber's) is plausible but unworkable. Training is driven by some recognition that something more than simple 'stick-and-rudder' and systems knowledge is needed. Because airlines, first, typically lack the skills of training analysis to work out their own solution and, second, do not want to do anything more than their competitors in order to meet a regulation we end up with the loose guidance we currently have in place. It is not the role of the regulator to tell an airline what the lesson objectives for a CRM class should be. The net result is as many trajectories and flavours as there are CRM facilitators. And you have to then think about local experience. The initial trigger for CRM was an aspect of civil aviation in the US (dominant captains, typically ex-military, who could not work as part of a team). In the early 1980's part of the debate was whether CRM was peculiar to the US experience and might not export. We can trace the problem right back to the failure to step back and take a broad view of the nature of aviation as a workplace. So, the CRM you have been exposed to is simply a product of the development of ideas.
As for measuring CRM, the problem here is that 'behaviour' is not a unit of output from a production process which can be checked against a standard. I can train and I can observe and I can give developmental feedback but we need to careful about expecting too much. So validation and repeatability are concepts that need to be handled with care in this context.
Your analysis of flight deck deviations merits more discussion than could be achieved in a single post but I'd take issue with your analysis. Skill-based errors might be increased through practice as I then place less attention on the task in question. I can execute the wrong skill because of a perceptual failure. Your decision-based error example is, in fact, a skill-based error. Decisions are all about selecting which skill sequence to use to get a job done. Violations are typically trade-offs between effort and risk. Interestingly, communication underpins all of this in that verbalisation of plans, current status, actual parameter values against expected values all can reduce deviations.
Last edited by turbocharged; 6th December 2008 at 09:54.