Book Review: Auditory Scene Analysis: The Perceptual Organization of Sound

The world is full of sources of sound. As I write this review, I can hear the humming of the word processor, the creaking of a door in the wind, the distant rumble of an aeroplane, the passage of a car close by, a bird twittering, my neighbour talking on his doorstep, music from his son’s hi-fi, and someone speaking on the radio in the next room. Although each source generates a particular pattern of changes in air-pressure, the changes have summed together by the time they reach my ears, yet I perceive each source distinctly. What principles of perceptual grouping and segregation do listeners use to partition such mixtures of sound? Which principles are applied automatically to all sounds? Which are specialized for particular classes of sound, such as speech? In what ways have the principles been exploited in musical composition? These are the major concerns of this lengthy, scholarly, but readable book. Bregman’s approach is functional not physiological, empirical not computational. He provides a comprehensive review and interpretation of perceptual experiments up to about 1989, so his book pre-dates recent attempts to implement auditory grouping principles in computational models and to find a physiological substrate for them. One important distinction is sustained throughout the book. Bregman argues that there are two kinds of principle for auditory grouping and segregation: “schema-based’’ and “primitive”. Schema-based principles are specific to particular types of source. They are learnt by listeners, and their application is under attentional control. One example may be the use of the knowledge of the timbre of an instrument to follow its part in an ensemble. Another example may be the use of phonetic knowledge to integrate acoustic cues in speech perception. Primitive grouping principles, in contrast, are innate, learnt through evolution. They automatically exploit fundamental physical properties of sounds and sound sources. For example: the sizes of resonators generally change slowly; they often generate energy simultaneously over a wide frequency range; when they vibrate, they create energy at the discrete