streaming of melodic patterns with high level cues?

[Greg Sandell <sandell@xxxxxxxxxxxxxxxxxxx>]

I wonder if someone can tell me if any research in the following area
has been done.

We know that interleaved melodic patterns will form separate streams
(and thus be recognizable) if they are presented at a fast enough rate,
and they are occupy separate pitch regions.  Furthermore, we can improve
on this (achieve streaming at slower rates and/or closer pitch
distances) if the two melodies are differentiated in some other way
(timbre differences, amplitude differences, envelope differences).  I'll
call this improvement "streaming enhancement."

Has anyone tried to find out if structure imposed *within* individual
melodic lines can help achieve streaming enhancement?  Here are a few
examples:
1. At each melodic jump in the melody (anything larger than a half or
whole step), make a timbre change, or have make the lept-to note louder
than its neighbors.
2. Use the same sort of cues (timbre or loudness) to indicate phrase
boundaries (assuming there is little controversy as to where these are in
the melodies).  For example, if the melody is "Happy birthday," there would
be a timbre change at each instance of the word "Happy".

These are only examples; I'm interesting in hearing about research that
employed *any* within-melody organization.

Thanks for any help,
Greg Sandell
--
Gregory J. Sandell  (sandell@epunix.sussex.ac.uk)
Lab. of Experimental Psychology, University of Sussex
Brighton BN1 9QG England  +44-273-678058