Randomizing identifiers, positions, and music prevents experimenter bias in listening tests

Evidence

Some audio journalists are hostile to the very notion that audio is amenable to scientific investigation—asserting that only subjective opinion, preferably theirs, matters. There is nothing wrong with subjective opinions. As will be seen, it is these that have allowed for deliberate and productive research into what matters in sound reproduction. However, the opinions must be collected under circumstances in which the ever-observant human brain is deprived of information that could bias the opinion about the sound. The tests must be blind—the listeners must not know prices, brands, sizes and so forth. Ideally they should be double-blind, so the experimenter cannot bias the results. There should also be comparison sounds from competing products readily available, because humans are notoriously forgetful when it comes to recalling the details of sound quality. When this is done the subjective data begins to resemble technical measurements, in that they become impressively repeatable and, more importantly, generalizable to a large population. — , Toole, Floyd E - Sound reproduction-…, 87 (#expands, via linked claim)

Categorization

• XREF: connects to: Experimental method determines what discrepancies listeners can perceive; Repeated familiar program excerpts make the most effective listening-test stimuli.; Blind tests collapse personal preference into shared judgments (+5 more)

Related

• Pilot tests and controlled constants eliminate confounding in listening experiments (similarity 0.94)
• Subjective listening tests require repetition and randomization to be statistically valid (similarity 0.90)
• Subjective tests have no right answers because each listener’s perception is unique (similarity 0.88)
• Every listening test has extraneous variables that must be controlled (similarity 0.88)
• Blind listening tests make subjective sound judgments objective and repeatable (similarity 0.88)
• Highly trained listeners can repeat subjective judgments with remarkable consistency (similarity 0.87)
• Subjective audio opinions can behave like facts when gathered under controlled conditions (similarity 0.87)
• Blind tests collapse personal preference into shared judgments (similarity 0.87)
• Simple forced-choice listening tests order noise samples by annoyance (similarity 0.85)
• Headphone blind tests cannot use automated shuffling, so tactile cues leak into results. (similarity 0.84)
• Blind listener tests reliably favor the most neutral loudspeakers (similarity 0.84)
• Randomizing positions and identifiers cancels room standing-wave biases in double-blind tests (similarity 0.84)
• Psychoacoustic knowledge rests entirely on subjective listening tests, not direct measurement (similarity 0.83)
• Subjective experience dictates which measurements matter in audio research (similarity 0.83)
• Repeated familiar program excerpts make the most effective listening-test stimuli. (similarity 0.83)

Broader Connections

• Blind listening tests make subjective sound judgments objective and repeatable #expands (0.78) — , Toole, Floyd E - Sound reproduction-…, 87
• Pilot tests and controlled constants eliminate confounding in listening experiments #contradicts (0.75)
• Louder playback is unconsciously rated as better quality in audio listening tests #contradicts (0.72)