Yes, and in addition to making sure the listener truly can’t tell which speaker they’re listening to via other cues (like audible position in the room, or sight of course), it’s also quite important that between each song test, you randomize which speaker you call “A” and which you call “B” to the listener. Otherwise, our brain (even subconsciously) will start forming a preferential bias that gets carried over from one song to the next.
Not shuffling A vs B label to the listener isn’t as bad as making the test “sighted”, but it ruins the statistical power of the test by making the whole test (no matter how many songs you have) no more statistically significant than if you had tested just a single song. By “resetting” the assignment of A vs B via random coin toss between each song (for example to decide which speaker to test first, then call the first speaker “A“ for that test), you assure that each subsequent song test is a new independent sample, which is very important and powerful.
Not shuffling A vs B label to the listener isn’t as bad as making the test “sighted”, but it ruins the statistical power of the test by making the whole test (no matter how many songs you have) no more statistically significant than if you had tested just a single song. By “resetting” the assignment of A vs B via random coin toss between each song (for example to decide which speaker to test first, then call the first speaker “A“ for that test), you assure that each subsequent song test is a new independent sample, which is very important and powerful.