6.3 Auditory Scene Analysis 101 understand the mathematical rules of the production mechanism so long as
they are relatively natural.
Continuity If natural sounds sources are limited by the rules of physics then we expect the
signals from them to behave in a constrained way. One such constraint is
conti- nuity , which implies some kind of low pass filter or slew rate limitation on how
fast some dimension of a sound trajectory may change. When it changes faster
than the allowed “speed limit” we tend to perceive this as a new sound, since
it would be impossible for the old sound to change in this way. An example
(by Warren 1972, 1982) reveals
homophonic continuity when a steady sound
is briefly increased in amplitude. Instead of hearing one object that quickly
moves closer and then recedes we tend to hear a new sound suddenly appear
in the foreground while the existing sound continues in the background at its
previous level. If the change is reasonably slow then it appears to be one source
that grows louder and then quieter.
Bregman found this applied strongly to localisation. If a sound equally
present in both ears was gradually increased in one ear, then the expected
perception of movement occurred. However, if the sound increased too rapidly
(less than 100ms) then listeners discarded this as an impossible event, since
the source would have to move too rapidly, and instead heard a second source
emerge in the direction of the louder sound while the original continued unal-
tered in the established position. The perception of this speed limit depends on
the perceived size of the source. Smaller, lighter objects such as a buzzing fly
can be presumed to change their behaviour rapidly, while larger heavier objects
seem to violate the perceptual speed limit for slower changes. This is reflected
in the wide range of continuity thresholds found for different sounds (Warren
1972; Bregman 1990; Darwin 2005), ranging from 40ms to 600ms.
Momentum We use the term
momentum here to distinguish it from continuity and mono-
tonicity. Although they are related concepts, broadly a
principle of good con- tinuity , they are not quite the same thing. Warren found that a continuous
pure tone interrupted by bursts of noise was heard as an unbroken stream with
bursts of noise added on top. Since the noise contains some of the frequency of
the pure tone it seems to be briefly masked by the noise rather than replaced
by it. Naturally, it is a more likely occurrence than the tone abruptly stopping,
being replaced by noise, and then resuming. It is as if the sound carries a weight
or momentum, making short interruptions seem less plausible.
Monotonicity Continuity of change can also apply to a sequence of obviously separate events
from the same source. Monotonicity is the tendency for an established pat-
tern to continue in the same direction of evolution. The interval between the
bounces of a ball always decreases. If the interval is small we might assume the
ball is very bouncy, but it never
increases unless some new event occurs to add