Thanks Peter ...

That takes my thoughts in a different and potentially more helpful direction. So it is not some meta-analysis of theoretical 'musical usefulness' of the phrase, but rather the number of times the phrase is utilized in the analyzed bars relative to other unique phrases that determines initial weighting. So then if a song, a dance song for example, has a fairly repetitive 'hook' on a rhythm track, the 'hook' phrase will end up with a very high weighting ... even though musically some of the other phrases could just as easily have been used instead for 'hooks' in the original composition.

In this case I see I need to learn how to better discriminate the bars selected for analysis or go back later and even out the cell weighting so that I can have the style generate greater compositional variety.

This suggests the analysis is done in the wizard 'bar by bar.' Is this true? There must be some way to isolate the phrase, and since midi data does identify each note by it's sequential position in a bar (sequence ... hmmm, wonder why that word sounds familiar?), I'm guessing that BIAB uses a 'bar' as the parameter box to analyze phrases.

Prado