A major question in motor systems neuroscience is how complex actions are encoded, on the timescale of both shorter individual elements and longer sequences. Birdsong is a tractable system, where a learned complex vocal behavior combines a categorical set of shorter individual elements into longer sequences, making it well-suited to address this question. The song nucleus HVC (used as a proper name) contributes to song sequence and timing. While much has been studied about HVC in zebra finches, which sing linear, stereotyped songs through their adult lives, relatively little data has been collected from HVC in their close relatives, Bengalese finches, that sing flexible, variably sequenced songs. We built a custom microscope to record neural activity reported by the calcium indicator GCaMP from populations of neurons from HVC in awake, freely moving Bengalese finches. We analyzed how populations of neurons in HVC encode information around divergence points, where one syllable can be followed by multiple syllables, and convergence points, where one syllable can be preceded by multiple syllables. We found that HVC projection neuron bursting can encode for upcoming sequence many syllables ahead of divergence points, and prior sequence many syllables after convergence points. We also found that HVC bursting encodes variation in the acoustic structure (phonology) of the different renditions of a given syllable in different contexts. Moreover, we found that HVC has overlapping representations of distinct syllables, especially those which are acoustically similar. These results help to reveal how premotor regions can encode multiple types of sequence and phonological information simultaneously.