As a pilot study into the understanding of the occurrence of extreme waves, the spatial development of an extreme wave (Ac/Hs=1.59) in a model basin was investigated. This wave occurred in a wave spectrum that was not extremely steep and non-linear. It is observed that the extreme wave develops in less than half the wavelength from a relatively normal wave into an extreme crest. The wave crest stays high and constant over a large distance (almost 75m). Linear dispersion is not able to predict the wave propagation towards the observed extreme wave crest. Second order theory improves the prediction of the crest amplitude, but not enough. The crest amplitude is still underestimated. This is confirmed by the plots of the probability of extremes. The linear Rayleigh distribution underestimates the crest amplitudes. The second order distribution follows the measurements much better, but also in this case typically the highest 10 crests in a 3 hours storm are underestimated.