Songbirds are an excellent model for exploring the neural coding of variable sequences of categorical acoustic elements. The domesticated canary, for instance, produce higly variable songs with complex transition rules between two consecutive acoustic elements. These transition rules are non-Markovian (i.e. the next acoustic element to be sung is dependent on several previous elements, not only the last one) [1]. In the HVC of Bengalese finches, Bouchard and Brainard [5] found that variations in responses to a given syllable could be explained by a positive linear dependence on the convergence probability of preceding sequences. Here, we reanalyse data from [3] to see if similar findings could be found for canaries.