Statistically significant poor performance (SSPP) in a listening test can result from either chance bad luck or experimental defects. Hence SSPP can be useful, but fallable, indicator of experimental defects. Defects that can produce SSPP include problems with research design, instrumentation, handling or analysis of data, and listeners. Statistical implications of analyzing results for SSPP are discussed, and a nearly forgotten procedure is recommended.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.