Geographic patterns of song variation are common in passerines and can develop as a consequence of the mechanisms of song acquisition and dispersal. In particular, the timing of dispersal relative to the sensory learning phase and the time of song crystallization is important. For example, when the sensory phase continues after dispersal or when males learn new songs every breeding season, i.e. open-ended learner, neighbourhoods can develop where males share their songs. Therefore, the analysis of song variation can give us some clues on bird behaviour. Using a comprehensive dataset of dispersal and song recording, I will present how machine learning approaches can be used to investigate the development of micro-geographic song variation in a wild passerine population. Automatic methods for song detection and comparison allow us to overcome perceptual bias in the classification of the songs of New Zealand hihi (Notiomystis cincta). We find that males tend to share more song elements of their repertoire with their neighbours than with more distant males or with males from the same natal area, implying that their repertoire is acquired post-dispersal.