The nervous system is made up of networks of many interacting elements at different scales. For example, in the human brain there are approximately a billion neurons connected by a million billion synapses. On a smaller scale, at each synapse there are hundreds of types of proteins which interact with each other.

Experiments can probe the function of individual elements in the nervous system and of behaviours of the whole nervous system. For example, much is known about how synapses change their strength in response to neural activity and about how animals remember. However, experiments do not explain how the high level behaviours result from the lower level interactions. To understand this, we need to reason about the behaviour of large numbers of connected elements. To do this rigorously we need mathematical or computational models.

In my view, the overall aim of computational neuroscience is to understand better how the nervous system develops and learns by constructing mathematical models of certain parts of the brain and central nervous system, analysing their behaviour either mathematically or using computer simulations and then comparing what the models do with what is known about what the nervous system does. The ultimate aim is for models that are consistent with experimental results, and which may make interesting predictions.