Nonlinear phenomena such as subharmonics, biphonation, or deterministic cha
os induce characteristic patterns in pitch- and amplitude contours. We intr
oduce a novel measure of correlations in contours termed MAC (Mean Average
Correlation). For uncorrelated and weakly correlated noise the Values of MA
C are calculated, and they turn out to be rather small. Moreover, we analyz
e simulations of the circle map and find high values of MAC for subharmonic
and chaotic data. We present examples of severely rough voices where corre
lations in contours are quantified by MAC. Finally, we study the connection
between perceived voice quality and MAC with the aid of 120 perceptually e
valuated voices.