Friday 25 July 2014

The Gold Standard?

Audiophiles often report being able to hear audible effects for which the currently accepted wisdom has no adequate explanation.  In such circumstances, the internet trolls - who are only ever skeptics - can be relied upon to invoke two ideas from medical science to disparage any viewpoints which disturb their comfort zones.  The first is the Double Blind Test (DBT).  According to those “experts”, unless you can demonstrate that a thing is audible via a scientifically-conducted DBT, then you have no basis for claiming that it is at all audible in the first place.  The second is that, when faced with a significant body of opinion that the disputed audibility is in fact real, they will attribute the result to “the placebo effect”, the strong implication being that idiot audiophiles are simply fooling themselves.  Unfortunately, the underlying basis of these invocations almost always goes unchallenged.  Is it right to uphold a DBT as the Gold Standard for determining subjective attributes?  And can the placebo effect be validly raised to explain away subjective observations?  I thought it would be worthwhile to take a closer look at those things.

A medical experiment often involves comparing one thing with another - usually a specific treatment and a placebo.  A DBT is conceived to eliminate two critical sources of bias from the test - those originating from the subject and those originating from the experimenter.  In the medical field, both sources of bias are unusually significant.  The subject clearly has a personal vested interest in the outcome of the test, and knowing whether they are receiving a real treatment or a placebo may compromise its validity.  A medical experimenter is typically a doctor, who has a overarching duty to the patient, and maybe also has a biased interest in the outcome of the experiment.  Bias on either part can destroy the validity of the experiment, and so a DBT is a test regimen which ensures that neither the subjects nor the experimenter are aware which of the test subjects are receiving which of the treatment options, knowledge which could affect the outcome of the experiment.

The first recorded DBTs were done during WWI with patients suffering from “shell shock”, a syndrome with which there was considerable disagreement within the medical profession regarding its interpretation, treatment, and prognosis - it was therefore particularly important that the treating physician be ‘blinded’ from the specifics of the ongoing test.  To be fair, DBTs are used these days in fields other than medicine (although much more rarely), but usually only where there is reason to suspect the influence of experimenter bias.

Incidentally, there is also a “Triple Blind Test” in which the person assessing the results and determining the formal outcome of the experiment is also blind to the details of the experiment.

A key element of a DBT is that the subject must be responsive to the test.  In other words, if we are testing the efficacy of an anti-cancer drug, it is no use using subjects who do not have cancer.  The responsiveness of the patient is usually assured by the fact that patient desires to be cured.  But if the test requires its subjects to walk five miles every day between visits to the clinic, it is important to weed out subjects who are likely to play fast and loose with their obligations to faithfully pursue the prescribed regimen.  Finally, if the outcomes are self-reported (for example, the subjects are asked whether their headaches disappeared), the experiment’s entire outcome relies on the subjects’ ability to self-report, both accurately and consistently.

So there we have the three pillars of a successful DBT.  Both the experimenters and the subjects must be free from bias, and the observed outcomes of the test must be both unambiguous and reliable.  That third requirement is one which high end audio’s internet trolls tend to forget.  It is no use whatsoever constructing a test with great detail paid to the blinding of the subjects and experimenters if the ability of the test methodology to delineate the anticipated outcomes has not been previously established.  In a medical example, this would be like using a DBT to determine whether a certain treatment made the subjects “feel better”.  The experiment would be totally pointless unless we had evidence to show that the methodology of the test can satisfactorily extract from the subjects reliable information as to the extent to which they “feel better”.

In audio, this third requirement is usually the killer.  ‘Blinding’ both the subjects and the experimenters is a trivial thing to do.  Extracting unambiguous test results from the subjects is not, and there are good reasons for this.  For a start, Joe Average is not a good test subject when it comes to detecting subtle differences in audio presentation.  The fact that Joe cannot perceive something is not proof that the thing cannot be perceived, even if you assemble a thousand Joes.  It is necessary to pre-qualify the test subjects, to establish that they can reliably detect the sort of differences that you are looking to test for under the conditions of your test methodology before you include them in the test.  The kind of skeptics who promote the concept of DBTs to resolve audiophile disputes usually find this requirement to be in some way unacceptable (of course), but that’s the truth of it.

Stereophile’s Michael Fremer - a favourite target of the internet trolls - has never made any secret of his willingness to participate in a properly conducted test to determine whether or not he can actually hear the things he claims he can.  He has even volunteered to go head-to-head with the Amazing Randi with $10,000 at stake.  But he has always insisted on having a veto over the methodology.  In other words, he won’t allow himself to be tested under conditions where he doesn’t think he will be able to perform.  He is very open about what he considers those conditions to be.  I find nothing wrong with that, and I fully accept that the more esoteric the differences he is expected to be able to resolve, the more picky his requirements are by necessity going to be.

The other medical tool in the audio skeptic’s armoury is the “Placebo Effect”.  Most people think they know what the placebo effect is, but very few actually do.  In reality, surprisingly little is known about it.  It arose from the first clinical drug trials, when it was discovered that patients tended to improve when treated with sugar pills which had no apparent active medicinal effect.  Such pills - and their procedural analogues - became known as placebos.  Therefore the standard that any drug or procedure had to meet before it could be introduced onto the market was that it had to outperform a placebo.  That is fair enough, but for decades the question was always “how can we make better drugs” and never “how can we make better placebos”.  That is changing, though.

A significant hindrance in pursuing placebos as a treatment was - and is - the troublesome ethical dimension to the debate.  Treating with a placebo would appear to require deliberate deception on the part of the doctor.  But placebo research is now a lot closer to mainstream, and the medical establishment is prepared to at least consider as valid a lot of the research into what the placebo effect actually is.  This research is beginning to deliver some unexpected answers, and while these are interesting they only serve to give rise to some new and unexpected questions.  For example, early studies appear to show that you can actually tell patients that they are taking placebos and yet still observe the same benefits.  Other studies are now looking into the physiological response of patients to placebo treatment, and comparing these responses to responses under other types of stimulus.  Just the simple fact of a doctor displaying a sympathetic and attentive bedside manner turns out to be an effective placebo, which may not be something your budget-constrained Minister of Health wants to hear.  So studies are also looking at the physiological responses of doctors while administering placebo treatments - who knows where that might lead!  With all this new interest, a lot of traditional non-western medicines - acupuncture would be a clear example - are now suggested as being perhaps elaborate protocols for administering the placebo effect.  It is a fascinating topic.  If you are interested to read more, do a Google search for “The Placebo Phenomenon” - the first hit should be from Harvard Magazine.  Read it.

The key takeaway for those who want to invoke the placebo effect in the audiophile debate is that it is a real and quantifiable effect.  Phenomena attributable to the placebo effect are undeniably real - it’s just that you don’t have an adequate explanation as to why.  It is wrong to invoke it in order to imply that something exists only in a person’s imagination, or is a product of self-delusion.  If a person insists that placing an Elvis Presley bobble-head on his amplifier improves the imaging depth of his system, it might be quite reasonable to be a little skeptical, but it would be incorrect to suggest that the placebo effect is responsible.  Actually, by suggesting it is a placebo effect, you are in effect acknowledging that the system’s imaging depth IS improved!  An Elvis bobble-head may be a lot of things, but a placebo probably ain’t one of ‘em.