Through a Glass Clearly
Of my many childhood memories of the Los Angeles Museum of Science and Industry, one particular exhibit sticks in my mind: a rectangular glass case about 4' long with many brightly colored geometric shapes suspended inside. The case is oriented so that the visitor approaches it from the side, wondering about the significance of this apparently random and meaningless jumble of colors and shapes.
Upon closer examination, the visitor sees an eyepiece at one end of the glass case. Looking through it, the incoherent collection of objects is revealed to form a complex and perfectly ordered pattern. The instant the display is seen from any angle except through the eyepiece, the coherence and meaning vanish.
I see a powerful metaphor in that exhibit for the way we experience music and judge reproduced audio quality. Just as one's perspective in viewing the geometric objects determines their meaning (or lack of it), one's perspective in music listening has a profound influence on the experience, especially in assigning musical significance to sonic differences between components. Point of view is everything.
This idea ties together several apparently disconnected views of audio and my recent experiences in auditioning components. One of these ideas was raised by John Atkinson in the introduction to his loudspeaker reviews in the January 1992 Stereophile, and expounded on in his February "As We See It." He made the point that, although we hear from loudspeakers distinct images in three-dimensional space, the two-dimensional audioband signals driving the loudspeakers contain no such images. The impression of acoustic objects hanging in space before us is purely a creation of the mind. JA asserted, therefore, that the nature of what a component does wrong is of greater subjective importance than the level of what it does wrong: "A large measurable difference may be inconsequential if it doesn't change the perception of [the] acoustic model."
The second idea is what Meridian's Bob Stuart calls "the increasing importance of the smaller difference." In my interview with him in last September's Stereophile, Bob related how on a particular CD, what sounded like a single guitar in the back of the soundstage was revealed by successive digital converter refinements to be first two guitars, then a guitar and a mandolin. The objective changes in the signal were quite small, but the subjective differencea single guitar becomes a guitar and a totally different instrumentwas huge.
I had a related experience listening to familiar music through the Mark Levinson No.30 Reference Digital Processor. The No.30's extraordinary resolving power, transparency, and dynamics produced a very different perception of the music. For example, the No.30's increased dynamic contrast and ability to present the steep leading edge of transients made drums more prominent and powerful. The result was a markedly different musical experience: the music had much more drive and rhythmic intensity, seemingly propelled by the drummer's energy. The entire band suddenly locked in rhythmically, with a greater sense of life and vitality when compared to other D/A converters. It's difficult to overstate the increased involvement in the music and the heightened sense of exhilaration.
The objective change in the signal that made the drums more prominent and lifelike must be minuscule, but the musical significance of that change was profound. Why? And how is the museum exhibit a metaphor for this experience?
In the museum display, the shapes and colors convey meaning only when viewed from a specific vantage point. Massively rearranging the shapes or substituting colors would result in no significant change when seen from the sidethe shapes and colors have no meaning before or after the rearrangement. But when viewed through the eyepiece, the pattern and coherence are destroyed by a tiny change in even a single shape's position. Similarly, small objective differences in a musical signal have a large subjective significance, but only when experienced from a particular perspective.
This perspective is an individual's sensitivity to sonic differences and how much that difference matters to him or her musically. The degrees of perception of a difference, and the ascribed importance of that difference, result partially from listening skill, but primarily from how the listener's feeling about the music changes. The greater the change in the value of the musical experience, the greater the perceived magnitude of that difference. The change has to have value to the listener to be perceived.
There is an argument that suggests that "subtle" refinements in audio reproduction are meaningless. It goes something like this: "Who cares about differences in cables, or even in amplifiers and D/A converters? Moving the loudspeakers 2" will produce far greater measurable differences in the sound." This sentiment was expressed to me recently by the editor of another magazine. He remarked that the same scale is used in describing differences between loudspeakers and amplifiers when, in his view, loudspeaker differences should be measured in miles, amplifier differences with a micrometer.
My thesis rejects this argument. First, no matter how the sound of a loudspeaker changes when it is moved in a room, its sound remains its own; it is not metamorphosed into that of a different model. More important, however, is that the kind of difference is more important than the magnitude of the difference. The relationship between objective magnitude and musical significance is far from linear. But what determines the linearity or non-linearity of the magnitude/significance relationship?
Perspective. How much the change is valueddetermined by how much one cares about the musicinfluences how much the change is experienced. To some, the change is not even experienced; others may perceive the change but not consider it meaningful; a few will find the difference staggering. But over this broad range of experience, which perception is true? Is there really that much difference between components? Are the differences described "true"? Our Western way of looking at the world implies that only one of the above perceptions can be true. I propose that there are as many "true" perceptions as there are listeners.
This idea, I believe, goes a long way toward explaining the conflict between so-called audio "subjectivists" and "objectivists." To the objectivists, who hear no differences between, say, loudspeaker cables, the subjectivists' descriptions of differences are either complete fantasy or magnified out of proportion. Whereas the caring listener looks through the eyepiece, the objectivists look at music reproduction from the side of the glass case. Changes in the geometric objects' position or color are imperceptible when they don't contribute to meaning and value. But from the eyepiece, even the smallest change has a profound impact on the pattern's order and meaning. The observer from the side can't imagine what the observer at the eyepiece is talking about; nothing has changed for him. It's no wonder that there's such a dichotomy between the two camps; both look at the same phenomenon, but from radically different perspectives.
This is one reason why blind listening tests are entirely inappropriate for discriminating differences between components; the procedure removes value and meaning from the musical experience. Such tests force the listener to view the music from the side of the glass case, rather than through the eyepiece. He becomes an observer of, rather than a participant in, the experience. If sonic differences have no musical value, they are not experienced. Consequently, any formalized approach to determine if differences existand weighing the significance of those differencesthat treats listeners and music as merely "subjects" and "stimuli," is doomed to failure. Further, this thesis explains how anecdotal listening (as practiced in product reviews) can reveal aspects of audio equipment quality not uncovered by formalized listening tests.
A paradigm of the failure of "scientific" listening tests is the case of Swedish Radio's subjective evaluation of low-bit-rate digital audio encoding/decoding systems. Low-bit-rate encoding (also called data compression) produces huge objective errors in the signalerrors reportedly masked by the desired, correctly coded signal. Consequently, the only way to evaluate such systems is through subjective listening.
Swedish Radio was given the task of designing and carrying out a listening test that would determine which of the competing low-bit-rate codecs (encoder/decoder systems) had the best sound quality. The secondand perhaps more importantpart of their charter was to determine whether one of the existing codecs had sufficiently good sound quality to become standardized as the replacement for AM and FM radio broadcasting in Europe. Clearly, this was not a trivial undertaking.
Swedish Radio's elaborate listening-test methodology was called "triple stimulus, hidden reference, double blind." Sixty "expert listeners" were used as subjects, with over 20,000 trials conducted on them. The listening-test conditions, methodology, and statistical analysis of the results were beyond scientific reproach. The entire formalized procedure was rational in every way. At the tests' conclusion, Swedish Radio proclaimed that two of the codecs "...had reached a level of performance where they fulfill the EBU [European Broadcast Union] requirements for a distribution codec." In other words, two codecs had, in Swedish Radio's view, good enough sound quality to become the digital replacement for broadcast radio in Europe (footnote 1).
Shortly thereafter, a Digital Audio Tape (DAT) of music that had been subjected to these low-bit-rate codecs was sent to Bart Locanthi, chairman of an AES ad hoc committee established to study low-bit-rate encoding. Within minutes of listening, he heard several problems, the most obvious of which was an encoder idle tone at 1.5kHz. He notified Swedish Radio of his observations; they reported having heard no such problems. When a few individuals at Swedish Radio listened again after Mr. Locanthi's report, they immediately heard the 1.5kHz idle tone.
This episode raises troubling questions. How could a single listener, under "uncontrolled," "unscientific," and anecdotal listening conditions, immediately detect an audible flaw that escaped the scrutiny of 60 "expert" listeners over the course of more than 20,000 "triple stimulus, hidden reference, double blind" trials? And with the future of European broadcasting at stake?
Perspective. That idle tone at 1.5kHz was like an additional geometric shape inserted in the glass case in the museum exhibit. Seen from any perspective except through the eyepiece, the additional shape has no meaning and is therefore inconsequential and escapes detection. But when viewed from a different vantage point, the additional shape completely destroys the order and symmetry of the pattern. Indeed, the intruding object comes to dominate the view.
What creates this difference in perception is value. We value the perfect geometric pattern seen through the eyepiece. We don't value an incoherent assemblage of shape and color. Similarly, what we value in audio components (and pay good money for) is their ability to convey the music's meaning and expression; not, for example, their low distortion or wide bandwidth. Technical specifications are like the geometric objects seen from the glass case's side; though ultimately contributing to the formation of the pattern, they must not be mistaken for the pattern itself.
This thesis raises questions about the transportability of subjective-review impressions to readers' experiences. If making value judgments about audio equipment quality is so dependent on perspective, how can I, as a reviewer, assume that the reader and potential purchaser of a piece of equipment will share my perspective and thus my value judgments?
The answer is that anyone who has expended the time and expense to read this magazine and others like it shares a broader, common perspective with the reviewer and every other audiophile: We all care about music and how well it is reproduced. Although we may have very different musical tastes and sonic priorities, the fundamental perspective of all audiophiles is identical.
We all look at music through the eyepiece in the glass case.Robert Harley
Footnote 1: See my report from the 10th International Audio Engineering Society Conference in last December's "Industry Update" for a more complete description of Swedish Radio's test methodology and results.