Reply to Bruno Repp (Yoshitaka Nakajima )

Subject: Reply to Bruno Repp
From:    Yoshitaka Nakajima  <nakajima(at)KYUSHU-ID.AC.JP>
Date:    Mon, 14 Mar 1994 14:06:14 JST

Dear Bruno,
     My former colleagues at Osaka University, Seiichiro Namba and Sonoko
Kuwano, have been interested in the problem of legato (in a broad sense) for
several years.  Their finding is known just among a few people unfortunately.
They presented this material in The 2nd International Conference on Music
Perception and Cognition, LA, in 1992.
     They took up the first bar of Moussorgsky's "Bilder einer Ausstellung"
played by a famous pianist, whose name I cannot spell -- Aschkenazy ?  The
melody goes: (g)(f)(b-flat)(c' f')(d').  The performed melody was clear and
'smooth'.  Then they cut out the first 200 ms of the second note (f), and a
terrible dissonance appeared, because the first note (g) still remained for a
while after the onset of the second note.  This demonstrates that there was
a considerable amount of overlap between the first two notes (g) and (f), which
was never perceived in the context of the melody.
     They are inclined to explain (as far as I know) that the remainder of (g)
after the onset of (f) is masked peripherally by the main part of the first
note (g) and the onset part of the second note (f).  But I am inclined to think
that the clear onsets of the notes facilitate the integration of a stream
(g)(f)(b-flat)---, and that the parts of the performed notes which are not
integrated easily into this stream do not contribute much to the perception of
this melody.  Because the listeners pay most of their attention to this melody,
the weak parts that are not related perceptually to this melody are easily
suppressed.  I cannot believe that the remainder of (g), which is so clearly
heard when presented with the first 200 ms of (f), is completely masked just
by adding the main part of (g), which would cause a forward masking.
     My former colleagues synthesized the melody using decaying tones, and
found that overlaps as long as 60 ms just made the melody smooth.  It seemed
essential to use decaying tones here.  A big question is, of course: Why do
the overlaps, which could cause dissonances, make the melody smoother ?  I am
very curious about the results of your new project.
                                              Best regards,
                                              Yoshitaka Nakajima