At least 16 non-expert listeners are needed for reliable speech-quality tests

Evidence

On each loudspeaker one set of technical measurements is sufficient, but subjective tests must be repeated and randomized, and must involve several listeners to be statistically useful data. — Toole, Floyd E, Sound reproduction- the acoustics and…, 144 (#expands, via linked claim)

Categorization

• corroborated: 2 independent passages support this claim
• XREF: connects to: Subjective speech ratings need at least sixteen listeners for statistical power; Subjective listening tests require repetition and randomization to be statistically valid — Toole, Floyd E, Sound reproduction- the acoustics and…, 144

Broader Connections

• Subjective speech ratings need at least sixteen listeners for statistical power #corroborates (1.00)
• Subjective listening tests require repetition and randomization to be statistically valid #expands (0.75) — Toole, Floyd E, Sound reproduction- the acoustics and…, 144