Greg's comment:

It's interesting to note that when we engage in a new culture, that in absorbing the practices and behaviours in that culture, we are, in effect, reinforcing those practices and behaviours.

Some of this culture can be adopted by a cultural entrant without their realising its significance or reason. The significance or reason may no longer be relevant!

This shared (redundant?) cultural knowledge has a binding effect on a community. They may even refer to it explicitly as underpinning their success. But cultural behaviour can also be detrimental to an organisation's goals and objectives. When this occurs, management may want to intervene and change the behaviours. But what about the virtual store of cultural knowledge? Can we really change that explicitly by changing behaviours or is the only way to replace it by firing a majority of people

One of the ways in which this (Cultural conformance) happens is though the absorption and repetition or personalisation of common stories. When you join a company, one of things that happens is that you hear what I sometimes call "foundation" stories, the ones that everyone knows and which constitute what this organisation is about. Until you can retell those stories in your own context, you are not really a part of the organisation.

To get really practical about this, one of the methods we created (Open Source and free, but this one is not documented yet so what follows is covered by a creative commons license) is to send people when they join an organisation on a treasure hunt. You give them some categories (A senior engineer with more than ten years experience, someone in accounts who has field experience) and tell them to gather in stories from those people. You don't give them names, they have to develop social networks to find them. Once they have gathered those stories, then, in front of their peers and after some training, they perform their own story, taking their own history, the stories of the elders and the current context to show how they stand in, not apart from the flow of history.

In effect that compresses around a year into 3-4 weeks. I got the idea from studying knowledge transfer in First Nation communities in Canada. Its a very different approach from trying to teach culture through explicit value and mission statements to which people can quickly learn linguistic conformance, but which they never really absorb.

I don't think you can change behaviour top down, but you can influence the evolution of narrative through shifting context. Firing people may be the only way in extremis but by taking a naturalistic approach, using the above and other methods you can both understand, monitor and influence the emergence of cultural change. But its bottom up, safe-fail influence its not design.