We found that across participants, the different conditions were characterized by differ-
ing degrees of phase coherence. Random sequences generated less phase coherence than all
other categories, and among the structured categories, scales (which showed the best phase
tracking) did notgenerate the highest coherence. Rather, 1/f^2 sequences were associated
with the greatest degree of phase coherence (Figure 21.9A). Interestingly, statistical research
on Western music suggests that melodies have approximately 1/f^2 statistics,61,62suggesting
that melody-like sequences generated more brain interactions than sequences which were
either overly random or overly deterministic.
To better understand the nature of these interactions, we examined topographic patterns
of phase coherence, subdividing the brain into four quadrants (anterior and posterior left
and right). We found that the greater phase coherence of 1/f^2 sequences was driven by
interactions between the left posterior hemisphere and the rest of the brain, including the
339Figure 21.8 Examples of phase-time contours (solid lines) extracted from brain signals recorded with MEG
sensors over auditory cortical regions. Each phase-time contour is overlaid on its associated pitch-time contour
(dashed lines: these contours have been down sampled to show the average pitch during each 2-s epoch in which
aSSR phase was measured). These contours represent single channel, single-trial data recorded from one listener.
Correlations between each phase-time contour and its associated down-sampled pitch-time contour are shown in
the inset to each graph. Full details of correlation between phase and pitch across sensors, conditions, and
subjects is given in Patel and Balaban.^44
Normalized pitch (dashed) and phase (solid)Time (s)Corr = 0.42Corr = 0.51Corr = 0.66Corr = 0.770 1 020304050600 1 020304050600 1 020304050600 1 02030405060BACD- 1
0- 1
0- 1
0- 1
01111