I have now read the entirety of this thread, as well as the "golden ears" welcome thread, so I'm going to make a comment. I think it's pretty clear that most of these people who come to discuss how they "hear things" with any of their equipment are trolls.
It is really quite baffling that people who will claim to "hear" differences between two identically measuring devices or cables can claim the high ground. The onus would be on them to show that there is either a measurable difference in the tested device (which Amir takes care of) or that there is some sort of measurable difference in the measuring device, i.e. their ears. Just telling people you can hear something different is absolutely not provable without a blind test. Of course this is all information that the non-troll members here already believe, but it strikes me how similar the concept of measuring and testing how humans perceive sound is to biomedical research and the never-ending quest to get rid of biases.
Regarding the placebo effect, this word gets tossed around a lot regarding audio sales and consumption. The placebo effect is well studied, however and is so important that the entire medical community considers this to be paramount in their research. There was a time when much of medical therapy was based on theory or "what works in my hands" but over time many treatments and medications thought to be effective have been proven to not be so. A common figure bandied about for the amount of effect that a placebo has is about 40%. But what does that 40% mean? It's not a plus/minus phenomenon. You can have patients that will have a placebo effect on top of an actual effect. That is why randomized, controlled, double blind clinical trials are done with medications. The goal is to eliminate bias.
There are some in the audio community that like to say there is no placebo effect with cancer drugs, or surgery, but this is not something that can be proven within the ethical constraints of modern medicine. You cannot ethically deny a treatment to a person, even if done in research, if there is a high likelihood of morbidity/mortality and the medication or treatment has a reasonable chance of working. You could only measure the effect of placebo on a surgery via a sham surgery, which just isn't going to be done on people outside of notably fascist regimes. When looking at a medication, a study may come back with a result of "works no better than placebo". This doesn't mean that you always get a placebo effect, though. Again, it's not a plus/minus thing, and some of that 40% mentioned before may not get as much "placebo" as another individual. It is likely that everything done in medicine has some form of placebo effect for some individuals but we probably will never know for many treatments. There is also a defined outcome measure that is easily and unambiguously measurable with these sorts of treatments- mortality. If it decreases mortality, it is deemed to be "working", but we still don't know if there is a placebo component here, it's just assumed there is not because of the gravity of the situation. But it's not been measured, and likely won't be measured.
There are three different general concepts when it comes to medical research for a medication. The first is mechanisms, meaning what does the tested drug actually do. You look at the molecular makeup, whether or not it blocks or up regulates an enzyme, occupies a receptor, binds a messenger ligand, etc. This is typically done in the laboratory where controls are easier to manage and will have results provided by instrumentation designed to measure all of those features to a highly detailed level. This would be akin to Amir's measurements, done on a benchtop in a controlled setting. And while this research is important, it is not going to tell you everything you need to know about how the medication will affect the body of a given human, or a group of humans. That is why we also have tests on pharmacodynamics and pharmacokinetics, how the treatment affects the body. These concepts tell us how a drug is metabolized in the body, what the distribution of the drug in the body will be, how long is the time to effect, how long effects last and those sorts of things. Then you have tests designed to measure how well a designed treatment or drug performs to achieve the desired outcome, i.e. measuring blood pressure, heart rate, blood cell counts, etc. You can also measure two of the senses with high precision, with field of vision testing and audiology.
Where things get really fuzzy with medical research is when the end result is not measurable with an objective test. There are times when the patient has to be the measuring instrument. The best example for this is probably pain therapy. For evaluation of how effective a pain medication is we have to rely on things such as a visual analog scale, which can be reproducible and used for populations, but is never going to be perfect. And you really can't take a large research study on pain medication and confidently relate the results to a single person. There are a lot of reasons why this is the case, but some of it can be ascribed to the placebo affect. There is also malingering and secondary gain, often seen with pain medication prescribing. There will always be that person who says, "yeah, I understand that the research says Ibuprofen and tylenol works as well as percocet, but only percocet works for me." There's just no way to really deal with that as a medical professional. And then there is also unethical prescribing, also seen with pain medication prescribing. This sort of research is really most prone to bias because of the patient bias. For a pain example, a patient may look at an unblinded option of extended release morphine versus high dose ibuprofen/tylenol, which may be equal on double blind studies, and say that obviously the extended release morphine will work better. But blinded the results are often different.
In summary, no ethical medical professional is going to go along with an unblinded study for medications or treatment. The medical world has known for decades that a randomized, controlled, double blind study is the only way to really determine if something works when it comes to the human body. There are certainly times when the best research isn't available and you have to fall back on theory and best experiences, but this isn't the case with audio. And so this seems to be where we are with the audio world when it comes to scientific accountability. Why would anyone believe an unblinded, non-controlled opinion on how well something sounds? Particularly, say, cables that cost thousands of dollars. Well, you have the placebo effect that is brought up here a lot. And it will undoubtedly explain most of the results people experience, and then there is the secondary gain of status. "Hey man, how does my $7,000 set of speaker cables sound? Good, right?". Thankfully nobody gets a buzz from snorting their copper. Is it possible that some people can have golden ears and hear something different for a given source? Absolutely. In fact, it is very likely that some do, if you look at probability curves. The numbers will be vanishingly low, however, and in the realm of statistically insignificant.