Dendrites are not static structures, new synaptic connections are established and old ones disappear. Moreover, it is now known that plasticity can vary with distance from the soma . Consequently it is of great interest to combine learning algorithms with spatially extended neuron models. In particular this may shed further light on the computational advantages of plastic dendrites, say for direction selectivity or coincidence detection. Direction selective neurons fire for one spatio-temporal input sequence on their dendritic tree but stay silent if the temporal order is reversed , whilst "coincidence-detectors" such as those in the auditory brainstem are known to make use of dendrites to detect temporal differences in sound arrival times between ears to an astounding accuracy . Here we develop one such combination of learning and dendritic dynamics by extending the "Spike-Diffuse-Spike"  framework of an active dendritic tree to incorporate both artificial (tempotron style ) and biological learning rules (STDP style ).