Subjective Fact or Objective Fantasy?

Blame the Puritans! say I. The high end has always had an ostinato accompaniment of grumbles from those who appear to feel that it is immoral to want to listen to music with as high a quality as possible. In a recent letter, for example, Fanfare and Stereo Review contributor and author Howard Ferstler states that "the audio world has more products of bogus quality and shills promoting them than any other industry, bar none," and trots out the old saw that audiophiles "end up spending an excessive amount of money on equipment or tweaking techniques of surprisingly dubious quality."

Mr. Ferstler goes on to say that the Stereophile of old was quaint, even enjoyable, and immediately recognizable as a religious, anti-scientific curiosity, though it did, and still does, pander to the desires of its readers, who are poorly educated in the sciences(footnote 1), to have a little mystery in their lives. He also implies that the current incarnation is less honest/more dangerous because it appears to be "scientific" and therefore possess greater apparent legitimacy. Stereophile and its writers, he says, "stand in relation to 'scientific' audio as creation scientists and their assorted publications do to legitimate studies of geology and anthropology...scientific half truths and misinterpretations are skillfully used to substantiate...religious views of the subject...contrary to scientific method, desired conclusions lead the search instead of a reasonably unbiased search leading to whatever conclusions might be revealed."

All very puritanical, yet I am disturbed by Mr. Ferstler's use of the qualifier "reasonably unbiased." This would seem to mean that he feels some degree of bias among those conducting an experiment is reasonable, perhaps even desirable. Scientific method is an unbiased attempt to describe things as they are, not how they ought to be. I fear that too often these days, practitioners of scientific method do indeed practice "reasonable bias," tailoring their work to produce desirable results. An editorial in the October 1988 Scientific American, for example, examined the fact that the findings of a committee on the nature of drug addiction in the US were rejected by the Reagan Administration as being too soft; ie, the scientists failed to prove that marijuana was as addictive as it was politically required to be. (Carcinogenic, yes; addictive, not very.) Similarly, scientifically trained John Sununu has been accused of editing official studies on global warming to render them more politically palatable.

The BBC's Hilary Lawson summed it up in 1985: "Science is there to be used, not to dictate what is true." The basis of scientific method is to look at how things are, then to design experiments to try to find the reasons why they are that way. In audio, many insist that such experiments are only valid when performed double-blind; ie, when neither experimenter nor subject is aware of the component under test. The problems with double-blind listening-test techniques are twofold: First, the subject does not judge the object under test directly—as in wine tasting, or in the testing of drugs—but only indirectly through its effect on an information-bearing, emotionally loaded stimulus—music. Second, the result of any scientific experiment can only be regarded as valid if all potentially misleading variables have been eliminated. This, of course, includes those introduced by the testing technique itself. As Robert Harley convincingly argued in his July 1990 "As We See It," the nature of listening under double-blind conditions is sufficiently different from the natural state of listening to music that results gained under those conditions are at worst meaningless, at best of limited transportability (footnote 2).

The proponents of double-blind testing in audio reveal their "reasonable bias" by performing a faulty experiment, noting that the outcome does not equate with their hitherto perceived reality, and therefore changing their perception of reality to match the experimental outcome. Such people use science as a drunkard uses a lamppost: for support rather than illumination (footnote 3).

By contrast, far from rejecting or perverting scientific method, Stereophile practices it in its true form: when experiments give results which contraindicate reality, the experiment is rejected, not the reality. Thus it is with amplifiers, for example: the results of many (but not all) blind tests notwithstanding, our continuing experience and that of our readers is that they sound different. Sometimes to a large degree, sometimes not so large, sometimes to an important degree, sometimes not so important: but different they do sound.

Why should their supposed sonic similarity be such an article of faith among "objectivists"? Amplifiers differ significantly in the way they measure, even on the small cocktail of measurements that Stereophile routinely performs, as exemplified by the VTL review in this issue. They differ in their ability to source current and therefore drive low-impedance loads. Their frequency responses can change at differing power levels, and in different ways. Their output impedances differ, producing differing response-modifying interactions with the loudspeakers to which they are hooked up, modifications that E. Brad Meyer admits in the June 1991 issue of Stereo Review are difficult to emulate with an equalizer (footnote 4).

The manners in which their transfer functions change with both output voltage and output current, hence the spectra of distortion products they produce, can be very different. The intrinsic distortion spectra themselves can be very different, tube amplifiers offering musically natural low harmonics, some solid-state amplifiers featuring musically objectionable higher harmonics. Some amps can deliver their specified power until the end of time; others have a power delivery that is correlated with the history of the music signal. Some introduce noise spikes when quiescent but not when delivering power into a load; some are only marginally stable and generate ultrasonic ringing on fast-risetime signals such as squarewaves; some will only demonstrate such instability with a particular combination of musical information and loudspeaker impedance; and so on and so forth.

Many prominent engineers—Stanley Lipshitz, for example—have stated that if two amplifiers are found to be audibly different, then simple linear errors, such as a frequency-response difference or an inversion of absolute polarity, will be the cause. Stereo Review's Executive Editor, Michael Riggs, went even further in the January 1989 issue of High Fidelity, stating that with the exception of loudspeakers, where it is still necessary to listen, "laboratory testing (properly done) can tell us pretty much everything we need to know about the performance of a typical piece of electronics...We know what the important characteristics are, how to measure them, and how to interpret the results."

With all due respect to Mr. Riggs, however, it is hard to see why "objectivity" should be the right tool to assess worth in an area which uses technology in the service of art. You can't measure the difference between a good and an inadequate performance of a piece of music, for example. As I write, I am listening to a recording of the Medici Quartet with John Bingham performing Elgar's Piano Quintet (Meridian CDE 84082). The music is immediately recognizable as Elgar, as late Elgar even, but what measurements could reveal that fact? Even if one used a computer to examine every word of the digital data describing both stereo channels of this work and how it related to every other word out of the total of 200 million, how could this determine that this was a turn-of-the-century piece of music? That it was by Elgar and not by Brahms or Dvorak? How, indeed, would the computer determine that this was a great piece of music and not just a similar-sounding potboiler? And that this was a very good performance, if not necessarily a great one?

I suggest that it is as much fantasy to expect that measurements can predict sound quality as it would be to think that they could predict the quality of music. All that the reductionism inherent in scientific method can do is to look after the fact for possible explanations for what is heard. And if amplifiers can sound different, it isn't necessary to invoke black magic, mysticism, or as-yet-unknown performance parameters as explanations. Architect Mies van der Rohe once said of his craft: "God is in the details." (footnote 5) When it comes to amplifier sound quality, the subjective differences are in the measured details. And the way in which to perceive how all those details interact and thus affect the sound of the music is simply to listen. Listening enables the whole of a component's performance to be examined simultaneously.

This does not mean examining every little aspect—How much bass? How liquid the midrange? How grain-free the highs?—but simply to sit, listen, and examine your whole reaction: Am I enjoying this? I believe this to be the basis of Ivor Tiefenbrun's famous "following the tune" criterion (footnote 6). Only by simplifying your mental activity as much as possible can you allow yourself to be truly receptive to what your senses tell you (something that students of Zen spend much time and effort trying to learn!).

"Quality" can only be inferred via a holistic approach, which is of necessity subjective. In fact, there is experimental evidence that the more a person tries to consciously analyze the available data before making a decision, the more likely he or she is to be plain wrong (footnote 7). If you want to know what a component sounds like, just listen. If the system doesn't detract from the music's emotional content, if it allows the music to "raise goosebumps," as Stereophile's founder J. Gordon Holt has phrased it on many an occasion, you're on to something good. As Corey Greenberg would put it, "The monkey bone doesn't lie!"

Footnote 1: Our 1988 survey revealed nearly 81% of Stereophile's readership to be college graduates, with 37% possessing postgraduate qualifications.

Footnote 2: Robert Harley will present a paper at the next Audio Engineering Society Convention—the convention's theme is "Audio Fact and Fantasy: Reckoning with the Realities"—to be held at the New York Hilton October 4–8. RH's paper takes a critical look at the objectivists' dependence on double-blind testing as the basis for their attacks on this and other magazines. It is significant that the two candidates for the 1992 Presidency of the AES are David Clark, who attacks Bob Harley in this issue's "Letters" column, and Floyd Toole of the Canadian NRC. Both favor blind testing techniques, yet to judge from their published work, the first exemplifies the convergent, closed-minded attitude to scientific research featuring a "reasonable" degree of bias, the second the divergent, unbiased, truly investigative, truly scientific attitude. I await with interest to see who the AES membership will pick in next month's election.

Footnote 3: "The patterns scientists observe in nature are intimately connected with the patterns of their minds; with their concepts, thoughts and values. Thus the scientific results they obtain and the technological applications they investigate will be conditioned by their frame of mind." Physicist Fritjof Capra in his 1982 book The Turning Point (Simon & Schuster).

Footnote 4: J. Gordon Holt discussed this interaction in Stereophile nearly 30 years ago, in Vol.1 No.5, May–June 1963.

Footnote 5: From which I conclude that Mies van der Rohe doesn't believe in God.

Footnote 6: The outrage engendered by Mr. Tiefenbrun's often hyperbolic but always perceptive statements gives me to think that the precepts they contest are not so much objective reality but rather articles of faith, culturally derived and therefore not open to question.

Footnote 7: "Thinking Too Much," Washington Post. Our thanks to Convergent Audio Technology's Ken Stevens for bringing this article to my attention.