MUSHRA only yields reliable results when sample differences are easily perceptible

Evidence

If the differences are small in the test cases, more accurate methods should be used, such as the method described in ITU-R BS.1116-1 (1997). … The number of samples is recommended to be less than 15 (including anchors and references) and the perceptual differences between the samples should not be too small, since otherwise the comparison may be too challenging. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 2354

Categorization

• XREF: Connects to psychometric limitations of rating scales — links loosely to how subjective scales saturate or discriminate poorly at small effect sizes.