May 1, 2009

The Misinformed Misleading the Uninformed -- A Bit About Blind Listening Tests

Let’s say I handed you a glass of wine or a soft drink and asked you to taste it, then tell me whether or not it was good. Assuming you could trust that I wasn’t going to poison you, we’d both know that it would be absurd if you replied, I can’t possibly tell you whether it’s good or not without first knowing who made it, what it is, and how much it costs. In audio reviewing, though, this happens all the time.

Not too long ago, The Abso!ute Sound’s editor, Robert Harley, wrote an editorial, "The Blind (Mis-)Leading the Blind," which has been reprinted on the magazine’s website. Read it and you’ll see that it is yet another article frowning on using blind tests in the reviewing of audio components. In particular, Harley says, "The answer is that blind listening tests fundamentally distort the listening process and are worthless in determining the audibility of a certain phenomenon." I feel he’s wrong, and although Harley encourages readers to participate in their forum, it seems more fitting that I write about it here.

Blind testing refers to the practice of concealing from the reviewer the identity of the product under test, in order to eliminate the bias associated with knowing the product’s make, model, price, appearance, etc. Blind testing is commonplace in everything from wine tasting to medical experiments; in the scientific community, it’s the only way tests can be done whose results can be accepted with any degree of credibility. You’d think it would be common among audio reviewers as well, but that’s not the case -- a situation reinforced by the kind of article written by Harley. Unfortunately, such articles are, in my opinion, examples of the misinformed misleading the uninformed.

Blind testing is a good way to reduce bias to make a more honest assessment, and has been used time and again to improve the audio products we use today. Blind tests are at the core of the decades’ worth of research into loudspeaker design done at Canada’s National Research Council (NRC). The NRC researchers knew that for their results to be credible within the scientific community and to have the most meaningful results, they had to eliminate bias, and blind testing was the only way to do so. Many of the companies -- Axiom, Energy, Mirage, Paradigm, PSB, Revel, etc. -- that participated in the NRC’s research achieved great success as a result, and use blind testing in their own processes of research and product development. Such firms know that researchers and reviewers aren’t the only ones susceptible to bias -- everyone in a company, especially designers, can succumb to bias and thus skew the results. Presumably, Robert Harley has reviewed products made by some of these companies. One has to ask: If blind testing is suitable for them, why not for him?

Probably the most eye-opening take on this comes from Harman International’s Sean Olive, who recently wrote about it in a blog entry titled "The Dishonesty of Sighted Listening Tests." Harman performed tests to see if there was a disparity in results between their blind and sighted tests, which Olive summed up as follows: "The psychological biases in the sighted tests were sufficiently strong that listeners were largely unresponsive to real changes in the sound quality caused by acoustical interactions between the loudspeaker, its position in the room, and the program material. In other words, sighted tests produce dishonest and unreliable measurements of how the product truly sounds." The idea that sighted tests, not blind tests, are highly unreliable should send shockwaves through the reviewing community, and lay articles like Harley’s flat on their backs. It should also have consumers eyeing with suspicion all product reviews based on sighted listening.

I’m biased toward blind listening tests because I know they work. I’ve participated in blind listening tests at the NRC, as well as at some of the manufacturers mentioned above. I and some reviewers have also set up blind experiments in my listening room to help us assess the performance of certain products. I find blind listening actually easier than sighted listening because I don’t have to concern myself with anything about the product other than its sound. Blind listening allows me to better focus on that sound. What’s more, there’s rarely a case where I can’t hear differences with the sound, which runs counter to Harley’s argument that blind testing distorts the listening process.

A component such as the Classé Audio CAP-2100 integrated amplifier helps makes the testing of source components and cables easy because you can match the levels of the various inputs, and turn off any inputs you’re not using. Provided the listener has not set up the system, and can’t see which source is hooked up to which input, the listener can then "blindly" select those inputs/sources, and then assess only what is actually heard.

That said, these days I’m part of a minority among audio reviewers. For a long time, I’ve wondered: Why are so many reviewers dismissive of blind testing and reluctant to have any part in it, particularly when it can be shown to be highly effective? After more than 13 years of reviewing, and of seeing what goes on in the reviewing community, I think it comes down to two things: a lack of knowledge and fear.

From what I can tell, those who dismiss blind testing have actually never participated in a well-designed blind listening test. When I hear reviewers dress the practice down, their knowledge seems to come from what they "know" of or presume about blind testing, not what they themselves have experienced of it. Therefore, I have to assume that they simply lack a clear understanding of how a blind test works. Perhaps if they had this knowledge, they’d know that, in a well-designed blind test, it’s quite easy to distinguish between products, provided there’s something to distinguish them.

Then there’s fear. It’s not farfetched to think that some reviewers’ "golden ears" may not seem so golden if it’s disclosed to readers that they have trouble arriving at the same conclusions under blind conditions that they do in a sighted test. Right now, reviewers are operating like card dealers who not only have the odds of the house on their side, but also the ability to see the cards before they’re dealt. I’m pretty sure that reviewers who are unsure of their ability to hear with only their ears have a vested interest in keeping blind tests away from their work, lest the world find out what their ears are really made of.

I believe that if those opposed to blind listening were privy to a well-set-up blind test that allowed them to listen at their leisure the way they do in sighted tests, the only difference being that they wouldn’t know the identity of the product they were listening to, the results might surprise them. I also believe that if blind testing were relied on more than sighted testing, it would make for fairer reviews and more useful results. After all, I’m sure that all reviewers -- even Robert Harley -- without knowing anything more about it, can tell you whether or not they like the taste of a drink. Why can’t they do the same with audio gear?

The downside: Although I believe in blind testing and the good it can bring, it’s not always practical to do, which is why you don’t see much of it in SoundStage! Network reviews. Next month, I’ll talk about the challenges involved in actually conducting blind tests, and what we’re attempting to do to overcome those challenges so that we can institute more such tests for future GoodSound! reviews.

. . . Doug Schneider