Auditory continuity and loudness computation

Sequences composed of alternating bursts of different levels with no silences separating them can give rise to a perception of a continuous sound upon which is superimposed an intermittent stream. These experiments sought to determine how the perceived loudness of the intermittent stream depends on the level difference between higher-level and lower-level bursts in the sequence in cases in which continuity is either heard or not heard. In the main experiment, listeners were asked to adjust the level of continuous or intermittent comparison sequences to match the loudness of components that appeared to be either continuous or intermittent in an alternating-level reference sequence, thus urging them to focus on the two-stream percept. Loudness matches of continuous comparison stimulus were close to physical levels of the lower-level bursts, whereas matches of the intermittent comparison stimulus were well below the physical levels of higher-level bursts. These results are discussed in terms of Bregman's [Auditory Scene Analysis (MIT, Cambridge, MA, 1990)] "old-plus-new" hypothesis: The loudness of the intermittent stream should result from the subtraction of the lower level from the higher level under the assumption that the higher-level burst represents a simultaneous mixture of sounds including the continuation of the lower-level burst. Additional experiments verified that, in the absence of the continuity phenomenon, matched levels were very close to the physical levels and that matches to fixed-level continuous and intermittent sequences were precise. The matching results from the main experiment support predictions of neither classical loudness models that do not take auditory organization processes into account nor schema-based models that presume a selection of information from the higher-level burst that does not affect the perceptual content of this burst. The matched levels fell between predictions of models based on subtraction of acoustic pressure and acoustic power, but were very different from subtraction of loudness measured in sones, suggesting that loudness is computed subsequent to auditory organization processes.