Auditory Masking and Signal Detection Theory

Abstract

The contributions of signal detection theory to the study of auditory masking are examined from two points of view. (1) Theory: In a given masking situation, the theory of signal detectability specifies the limitation imposed upon the performance of an observer by the characteristics of the signal and of the masking noise. The limitation upon performance imposed by a particular environment is represented by the limiting ROC. The theory of limiting ROCs is the theory of ideal observers. Each masking situation determines a particular ideal observer. Factors that degrade performance relative to the appropriate ideal observer, such as filter bandwidth and methods of processing, may then be investigated. (2) Methods: The fundamental detection experiment, that is, the single-interval paradigm with a yes-no response, must provide data for the estimation of two, formally independent, conditional probabilities. One probability is the hit rate, and the other probability is the false-alarm rate. These probabilities determine a point in ROC space. Both probabilities must be estimated, because the decisions of the real observer may be influenced by nonsensory factors that are not subject to direct experimental control. The advantages of a two-interval task are discussed.

Key Words:

• Statistical decision theory
• Receiver Operating Characteristic (ROC)
• False-alarm rate
• Hit rate
• Psychometric functions
• Two-interval task