Neuron modelling is faced with challenging hurdles. Both realistic and simplified models are often high dimensional — that is, they have many unknown parameters. This can lead to model overfitting, which makes it difficult to assess the correctness of the model. Additionally, good models should provide predictions that can be refuted experimentally. Ensuring compatibility with known experimental results is easier than predicting new, yet-to-be observed phenomena. This talk introduces a theory that offers a new prediction about the response of sensory neurons that is not dependent on the selection of parameters. This allows both the model and the prediction to be experimentally tested without the need for selection or fine-tuning of parameters.