The Korf Blog

The inside story: our research, development and opinions

Previous
Next
12 September 2017
Validity of ABX Testing

Here are 3 pairs of adjacent squares. Some are the same colour, some are different. If you have a half-decent monitor and no significant vision impairments, you should have no difficulty telling them apart.
Blues are the same, while greens and reds are subtly different. Subtly, but quite visibly.
Now let's change the task a little bit. Look at the first series of squares, and then scroll down to see the second one.
Keep scrolling...
Keep scrolling...
Keep scrolling...
Keep scrolling...
Now try telling: which ones are different this time? Scroll down for an answer.
Keep scrolling...
Keep scrolling...
Only the brown ones are different.

What makes the second test much harder is the fact that in the first test we're testing perception, and in the second one retention. To succeed in latter, you have to have colour memory, not just colour perception.

And this is the problem with ABX testing in audio. There is no reasonable way to present 2 or more audio streams simultaneously. Thus, all ABX protocols rely on sequential presentation of recordings. They test audio retention instead of perception, and aural retention in humans is notoriously weak. 1/10th of a second pause, and we cannot tell similar sounds apart at all.
Don't forget to subscribe to get the latest blog updates!
comments powered by HyperComments