ABSTRACT
The appropriate level of detail for a computational neuroscience model is determined by the nature of the system or phenomenon under investigation, by the amount of experimental data available and by the aims of the investigator. For example, models of certain network phenomena, e.g., synchronization, do not require details of cell morphology, and perhaps not even ionic currents-integrate and fire neurons will suffice. On the other hand, studying dendritic processing in active dendrites requires details of dendritic morphology and of the active channels in the dendrites. However, although a detailed model may be desirable, such a model must be well constrained 1-58488-362-6/04/$0.00+$ 1.50
by experimental data or else be over-parameterised and suffer from lack of predictive or explanatory power. The final factor affecting the level of detail is the purpose of the model: to provide support for a specific hypothesis about the function of a neural system-such models tend to be more abstract with less detail-or to be guided by experimental data to discover unknown properties of a system-such models tend to be more detailed, aiming at as realistic a model as possible.
