One of the many items lodged in my musical memory (and it’s been there for forty years) is the ‘ritornello’ heard at the beginning of Bach’s Fifth Brandenburg Concerto, all 121 notes of it. You can see it notated here. (I didn’t have it quite perfect, it turns out; I thought the eighth-note ‘e’ at the end of the second bar was two sixteenth-notes.) I can play it through in my ‘mind’s ear’ at will, or sing it (though the range, of over 2 octaves, makes that a bit tricky). It’s not ‘random-access’: I basically have to start from the beginning.
Now, to represent this tune in a straightforward computer-friendly way would occupy about 4096 bits. But, obviously, its brain-based representation is nothing like what a computer would use.
In one sense there is far less information content than the note-for-note calculation would imply: the tune is ‘memorable’ precisely because it is a tune; it makes musical sense. An arbitrary sequence of 121 notes chosen from a span of 2.1 octaves would be virtually impossible to memorize. In a real tune, the choices for each successive note are limited; we could draw an analogy with the predictive algorithms by which the email app on your phone offers you a few likely choices for each word in your message based on what has gone before.
But consider what this means for the brain-based representation of the tune: it is tied in subtle ways to a whole subsystem of musical knowledge.
Just ponder the implications for the gap between the fact-of-the-matter about connections between neurons and the meaning that those connections represent. More to come on this topic.