To discuss the governance of innovation in animal production three innovation models are placed in the context of the phases of development of agriculture (according to Hurst, 1997, Crisis and renewal - meeting the challenge of organizational change. Scriptum, Schiedam). The phases distinguished are spontaneous action (breakthrough of a new paradigm, rational action (heyday) and action under restrictions (new choices for science, business and public). The associated innovation models are the Participatory Technology Development (PTD) model, the linear model and the chain link model. It is argued that the linear model has been the predominant one in the past half century where food security was the prime drive for action. In the last decade this drive clearly fades away and new goals of animal production emerge, requiring another innovation model. Using the examples of two firms it is illustrated that the chain link model, along with the linear model seems an efficient way to deal with changing circumstances. To show the dynamics of the system, the model is extended into one in which the three models of innovation do not follow after each other (with the chain model as the end model), but where both PTD as the chain link model are starting points in situations of change to take aspects of the linear model on board when heyday emerges. It is argued that in the more dynamic context where heyday clearly is not predominant another role of researchers is required. Where during heyday, participation in the process of optimisation and fine-tuning of production systems is a successful approach, during the other two phases a problem observation role is required, where the researcher takes part in the public debate on direction and usefulness of solutions. (C) 2001 Published by Elsevier Science B.V.