HERE GOES ANOTHER VERY LONG POST. I AM SORRY IN ADVANCE.
1.0 Why am I here and what the f**k do I know?
Something I dislike: coming to a new poster board and seeming like I am some pedantic asshole who thinks he knows everything about everything. I know so very little about so very much and I am fully aware of this. I came to SPCR to learn about computers, something I know almost nothing about, but through a comedy of errors have ended up trapped in a few threads that have ended up dealing with one or two of the only topics in the world I do happen to know something about: audio equipment and scientific method/testing methodology. (And I don’t consider myself an expert in the former, I personally know a number of people far more knowledgeable on this topic than I—I have never even designed a circuit or built a speaker in my life, how much could I really know?!?!) Weird. That said…
2.0 Reality and perception
2.1 Rationalists and … irrationalists?
Devonavar, and others, have suggested that there is a divide between “rationalists” and “subjectivists” within the audio community. I reject his characterization of people like me as being non-rational, for reasons I shall detail below. On the other hand, I think he brings up a good point when he uses these themes to address the question of what is the standard of “real” in an medium that is inherently divorced from the empiria. Contrary to the prevailing view around here, audiophiles are not a bunch of gullible sots, forever chasing the next brand of “snake oil” (though some are), and this very topic has been the subject of long and heated debate in such places as www.audiogon.com
, as well as in Stereophile and The Absolute Sound (TAS). This topic cuts to the heart of a matter that is both ontological and epistemological in its nature.
Ontology is the nature of things. Subjectivists (e.g. post-modernists and some social constructivists) argue that “reality” is contingent on those observing it and in discussing what a sound really sounds like in abstraction from the person doing the listening is an untenable exercise. Their cause really got underway after the failure of Aristotelian Physics, which everyone just knew was “true” after centuries of testing, in favor of Einsteinian Physics. That is, they claim that Aristotle was “right” only because so many people happened to believe he was—and that he had now fallen from favor was simply a result of a shift in prevailing views, not “truth” per se, which does not exist outside of the eye of the beholder.
Objectivists argue that an objective reality exists—e.g. a reality that is an object separable from the person doing the observing. One need not abandon rationality to admit that this reality may be only imperfectly perceived and doing so does not deny its (meta-)physical existence. Perception for rationalists becomes an issue of epistemology—or how we “know” the things we “know.”
So how do we know what we know? Soft-core social constructivists (those not associated with ontological subjectivity), argue that even if an independent metaphysical reality exists, our perception of it is so colored by our socially constructed identities (which encompass gendered, racial, social, familial, etc. social norms, rules, and relations), that an objective observation of a “real” reality is not possible. Rationalists, on the other hand, believe that our perceptions (and tests, et al.) do allow us insight into the “real” reality. The “noise” associated with these perceptions will dictate the degree to which we correctly perceive reality and thus the proximity of our beliefs to it. The greater the “hard” rationality assumption, the lesser the assumed noise in estimation; in the limit, noise --> 0 and perception = reality.
2.3 Political and the social sciences
Last year, I received my Ph.D. in political science from a department that is well known for its rational-choice approach to the field—from both an ontological and epistemic perspective. That is, we are working very hard to put the “SCIENCE” back into political science. Game theorists and quantitative methodologists abounded and, perhaps because I am a product of this factory, I have bought into this program and am on rat-choice bandwagon in a big way.
I am convinced that a more rigorous, scientific approach to political science (and the social sciences more generally) will yield great insights. On the other hand, I also know that we do not have access to all of the necessary data, models, and testing techniques we would need to fully (epistemologically) specify the world according to these standards. Indeed, often times we must move our analyses from the quantifiable world of data-dominance and address many substantively important issues using impressionistic, descriptive methods. If we were to limit ourselves to questions upon which we had fully specified methods and data, I fear we would remain unable to address far too much of what is truly important in the world. Thus we must, at times, go beyond quantitative “tests” knowing that though we may draw conclusions when doing so, these conclusions remain suspect and open to revision.
In a circumstance such as this, how are we to incorporate new data, be it quantitative or qualitative, into our belief-set? Bayes rule offers us one possibility. Bayes rule, simply stated is: Pr(A|D)=Pr(D|A)*Pr(A)/ Pr(D|A)*Pr(A)+ Pr(D|~A)*Pr(~A), which we might read as something like: “The probability A is true after observing data D is a function of the probability we would observe D if A were true, times our prior belief that A was true, divided by the probability we would observe D if A were true, times our prior belief that A was true, plus the probability we would observe D if A were not true, times our prior belief that A was not true.” In effect, Bayes Rule gives us a ratio into which we can incorporate incoming data in a “rational” manner. In the absence of spiked priors (in which all the probability weight is distributed to our prior beliefs and zero-weights are given to new data), we have a means by which rational observers can update their beliefs such that as D --> (infinity), our beliefs --> reality.
This is a very powerful rule because it says that all individuals, WHATEVER THEIR PRIOR BELIEFS AND NO MATTER WHAT WEIGHT THEY GIVE INCOMING DATA (AS LONG AS IT IS > 0), THEY WILL ALL REACH THE SAME CONCLUSION AS D --> (infinity). Not only that, but that conclusion with be an epistemologically “correct” belief over the true ontological nature of the empiria. Note that Bayesians are thus, by their nature, both ontological objectivists and epistemological rationalists. However, note that if D < (infinity) we can only approach truth (at varying rates of closure)—thus even rationalists will admit that, in the absence of full data, we may still hold beliefs that only approach (or approximate) reality.
3.0 What does this have to do with audio?!?
3.1 Bayes rule and the audible phenomena
I said above that I disagreed with Devonavar’s characterization of me as non-rational. I am fully rational in my outlook in a Bayesian sense: I hold certain beliefs and test them against data and update my beliefs at a rate that suits me. I also believe that musical events are ontologically separable from those listening to them and that there IS a “true” sound for a given performance.
How we can know what that sound is a thornier question. As has been noted by others, the most important issue in determining what we hear is not made by us at all, but by record producers and recording engineers (for a very insightful discussion on this point, see the Round Table discussion on the playback hierarchy in TAS no. 151: pp. 84-93). They are sure to screw this up… but hopefully not too much. We have no control over this process, so the best we can do is select components that are most faithful to the recording and set them (and our rooms) up so that they will most accurately reproduce the information held in our aluminum pits or vinyl groves.
But is it possible to know if we are getting closer to (recorded) reality? Well, as a rationalist, I would say yes. We have a great deal of data upon which to draw. A FIRST CUT would be electronic testing. This is an excellent tool, one that is completely scientific and replicable. Unfortunately, as in political science, we have access to neither the full data nor the fully specified tests necessary for this first step to be enough on its own. If we want to answer many of the interesting questions, we must UNFORTUNATELY move beyond these methods and on to others that remain more impressionistic and descriptive… and, yes, experiential. We can each, individually weight the degree to which these new (qualitative and quantitative) data revise our posterior beliefs, but to disregard either (as many posters here seem to do) or both is as foolish as it is inefficient.
Note that, by updating my beliefs according to Bayes Rule, I am assuming that (1) incoming information is a true (if noisy) signal of the empiria and, as a result, (2) updating allows us to converge toward true belief. However, given that D < (infinity), we cannot say that we have reached the full epistemological truth content of the audio playback phenomena. However, because I do not spike my prior beliefs and do allow for updating, I am certainly more rational in my methods than, say, yeha.
3.2 But if “correct” remains unknown, should we prefer “pleasant” instead?
Again provocatively, Devonavar asks that if we cannot know for certain when the last bit of information is extracted from a vinyl groove or recovered CD pit, might we not seek a pleasant musical experience as opposed to a “realistic” one? I think not, and I think most audiophiles would agree with me. As we love music, there can be nothing more appealing then reproducing the musical experience as faithfully as possible—subject to the constraints placed upon that experience by the recording side of the equation. It is true that certain gear adds something that may correct or enhance certain recordings or types of music, but often this is simply a band-aid for failings that should have been corrected in the recording studio—our playback rigs cannot be held hostage to a few poor recordings—and the more faithful our playback gear, the better is will sound on the widest range of material (although, again, it might sound down right terrible on a few select pieces). (The problem is that both the audiophile who listens to 3 watt SET amps and Lowther speakers and the other audiophile who listens to Krell mega-amps and difficult-to-drive cone drivers both say they are after truth… and one HAS TO BE WRONG!)
How can we know if we are moving in the direction of greater truth as long as the full measure of truth eludes us? Bayes rule provides us a way: if the empiria is a true (if noisy) signal of truth and if we update our beliefs and expose ourselves to as much and as varied data as possible, our beliefs over which components are more truthful will converge towards reality—even if it never reaches it.
We will be confident in the incoming data, and can then update at a higher rate, the higher the (perceived) quality of these data. For example, I hold lab test results in a somewhat lower regard than many here so I am less convinced by lab results telling us that there is no difference between a $$$ CD player and a Wal-Mart Special, but new and more refined electronic testing techniques may increase my faith in what our instruments tell us. Also not that, skeptical as I am, I do not disregard these results, indeed, I would begin to regard the expensive CD player with some suspicion after observing these results, I just to not allow them to update my posterior beliefs to the same extent as, say, yeha. By the same token, training our ears is always a good idea. Perhaps if yeha were to attend more live concerts (particularly of acoustic music), he might have enough faith in his own ears that he would be willing to let them judge the differences between that same $$$ CD palyer and the Wal-Mart Special (not as opposed to lab measurements, but in addition to them). Note that these results only obtain if we begin by assuming ontological objectivity and epistemological rationality, however.
4.1 As much as we can, as often as we can
Bobdog : How can you buy shakti stones and challenge yeha's stats and numbers?
I cannot understand this statement at all. I buy things like shakti stones in order to TEST them! I try and buy, try, listen to, and sample, as much live and recorded music and playback equipment as I can—some of it is quite goofy. I don’t consider this irrational, but rather an opportunity to update my beliefs over certain technologies and products.
Consider this: all rational scientists (social or from the “hard” sciences) are at heart Lakatosians; Imre Lakatos’s work on the philosophy of science is a corner-stone in the scientific method. He (and Popper before him) proposed that we made BOLD predictions (like “shakti stones will work!”), and then subject these hypotheses to hard empirical tests (as this post is too long already, I will avoid the finer points of Lakatosian research programs). In this vein, the more strange the claim, the more potentially informative will the tests on them be because it is these bold, counter-intuitive claims that most challenge our comfortable (and often incorrect) presumptions. Although they are often proven false, such claims are also the path to real breakthroughs as well—to shy away from these, as tay suggests, is as unscientific an endeavor as I could think of.
But what do these new claims do to our existing beliefs? How do we incorporate the data we have from or tests of the same? BAYES RULE PROVIDES THE (RATIONAL) ANSWER!4.2 Blind testing
One final note: Many have advocated double-blind testing to see “if there really is anything to this ‘hi-end’ audio stuff.” I have as well—I think it is a source of great empirical data (though not the only one). The reason why some audiophiles are more leery is this: some gear is made to sound/look good on a short demo but is crap in reality. TV’s factory setting generally crush white levels and make everything look blue in an effort to get that TV to “pop” on the shelf—providing a brighter, if not better, picture than anything else near it; by the same token, some audio gear will be pumped up, punchy, bright, and in-your-face which will surely catch your attention on a 5 min. demo in the store or on a 30 sec. cut in an ABX double-blind test. However, like bright blue TVs, these characteristics will grate in the long run (I think of Polk speakers when I think of this).
Many audiophiles fear that such jacked up gear will get too great an edge over more subtle equipment that relies on uncovering nuance and detail that will reward a long-term listener but will be overwhelmed by bombast in a short audition. On this point, I actually agree with yeha, somewhat: I think this is no excuse for abandoning double-blind testing. But I also recognize that sometimes longer experience is needed as well. There are many tests we can use to evaluate audio gear; it would be great if one day we had a set of measurements that would save us the trouble of listening to see which products were best (… I guess), but we are NOT THERE YET. As a result, we must avail ourselves of a number of testing methods, some scientific (lab tests), some quantitative (properly run ABX tests), some personal and experiential (do I actually hear a difference?), and they should ALL be allowed to inform our posterior beliefs. Given our limited understanding of the audio (psycho) phenomena, it is the only rational thing to do.
And if this post has contributed little to the overall thread,... then I am apologetic but I'm glad I'm in the overall majority...
I liked that so much I decided to steal it.