Brooko's primary argument, that if it isn't measurable--by his equipment, nonetheless--it doesn't exist is a fallacy as I've previously stated. Absence of evidence is not evidence of absence. Brooko's small collection of measurements on a variety of headphones with equipment that is not designed to discern minute differences, is no better than throwing forward completely subjective arguments. I provided an example of a rigorous systematic review that found that measurement instruments in quality of life did not capture elements of people's experience that subjective interviews consistently captured across numerous studies in a wide variety of geographic contexts. By Brooko's line of reasoning, I should have ignored the subjective data.
Again - you are jumping to huge conclusions here - and I wish you would stop. I will state again, I am open to the fact that break-in could exist, but failing conclusive evidence my findings so far is that as far as true audible results go, it does not have the effect that others propose. I haven't heard it, I know enough about auditory memory to know that subjective haring tests are unreliable, and my proposition is that if burn-in does give the results that people calim - big changes to bass and treble - then it would show on a frequency plot. If things like changes in position of headphones on a persons head, or depth / angle of insertion for IEMs can make audible differences and clearly be measured - why would break-in be different?
Question - are you saying audible break-in would not show up in a frequency plot
Quote:
I have argued that it may be possible to prove or disprove burn-in, but we need sensitive enough instruments to do it, which neither of us have. Asking me to produce measurements trying to dictate what that the heart of the argument is about objective measurements. What we are actually debating is whether those objective measurement tell the whole story. So, asking me to produce measurements is like saying I'm right because I don't believe in your premise for argument and I won't deign to look at alternative evidence. I've already acknowledged that the information we have available in measurable differences is not enough to support burn-in effects, but I don't think that we should ignore qualitative data, or that we should assume that our instruments are perfect instruments. Brooko's request is like saying prove me wrong using only the evidence that I want to hear. That's a bull-headed way to look at it, and I should know, because I'm frequently accused of being bull-headed.
I've covered the equipment already - and look forward to Luke's reply. Given the magnitude of the claims regarding burn-in (easily audible, big changes etc) - I find it unlikely that this would not show on a frequency plot.
You keep coming back to the "well I heard it" type of debate - and when presented with measurements, the standard response is that they aren't accurate enough. I'd like you to remember this point as I'll cover it in my final post.
Quote:
I don't know how I'm supposed to debate someone who doesn't have any desire to do anything to counter my conclusion about temperance of conclusions and how people make decisions under uncertainty. Brooko has yet to provide a rigorous study of this topic, but has basically told me that I lose the debate if I can't.
How can you make a decision under uncertainty? You don't have the facts. All you can do is make a guess. If thats what you have - and you choose to discard anything else - then I agree, this thread is pointless - and I'm willing to disengage now. The real fact is that you bring nothing new - just a return to the old "I heard it therefore it must be real".
I provided the example of equivalence trials, which is what would be necessary to prove that people can't actually tell the difference. Further than this you would need these trials to have a cross-over design, which is even further complicated. People would have to random order of headphone listening. There would have to be sufficient headphones to minimise the impact of confounders due to variability in manufacturing. Also, because we need to find out if people are just guessing, we need to have people repeat the evaluation multiple times. For validity in telling if someone can tell differences between how music sounds, multiple music samples would need to be available. Some of this material will be familiar with proponents of ABX testing, of which Brooko is one--a topic for another day or perhaps never. Of these elements of the study design, two reduce sample size: cross-over design, and repeat evaluations. All other factors make it very difficult and expensive to conduct the research. Here's some confounders you would have to adjust for additionally, listener expertise, age, and sex. Your randomisation could take care of those. To ask me to produce this evidence is unreasonable.
Why would we need any of this? Can we agree if its audible, its also measurable? And also that if its clearly audible (again get to that shortly) then it should show up on a frequency plot? If we can get to this point, then studies and trials you proposed aren't needed.
Brooko has conveniently presented the conclusion of one of Tyll's posts, whilst ignoring the other. Tyll conducted a single-blind experiment wherein he had to identify which headphone he was listening to, the one that was burned in, or not. He identified 13 of 15, and then 5 of 5. That isn't likely to be a result that occurs by chance. That experiment had limitations: it was single blind, so it is possible that the other participant could have somehow clued Tyll into the headphone he was using, it is unclear whether pad wear was controlled for (I doubt it), there were not enough headphones, observers or observations. As I said in my first post responding to Brooko, there are too many confounders to be conclusive. Did Tyll hear a difference, yes, can we make any reliable conclusion on it, no. In this situation, whatever our previous knowledge is will dominate, under a Bayesian perspective. This is why neither Brooko, nor I are convinced to change our position.
I actually find it funny that you presented your finding from the Matrix-Hifi post, claimed it was unlikely I'd read it, and then also put this bit about Tyll's second link - and claim that I'm ignoring it. I read both. I've read all of Tylls finding on burn-in (all the ones I know about anyway), because it interests me. I was hoping I would find something more- but all it has done is reinforced my own personal beliefs.
Lets look at the Matrix HiFi link. First of all - it is all regarding speakers - not headphones, not earphones / IEMs. By design they are different. Also - the largest portion of the manufacturers polled (over 1/3) all agreed that any significant burn in would happen in the first few minutes. Apply that to your observances of needing many hours for much smaller drivers - and any change not being noticeable until that time. Yes - I read the entire thing (twice now). I still say it has no relevance to our discussion. How many earphones do you know with a Spider and paper surround by the way? And the burn in on the Sirius - are you really trying to say that MAtrix HiFi link has anything to contribute at all when you are talking about vast differences in driver design and material?
Going back to Tylls second link - and it was actually posted 4 months before the one with the measurements - not afterwards. this is an important point I think. To do this experiment he used two headphones. Same manufacturer - same model - two headphones. He did not measure both pairs before hand to see the difference in frequency. this is important. Here is a snipped from his conclusion at the time.
What Does It Mean?
Have we absolutely proven that break-in is an audible phenomenon? No. All I've proven is that I could tell one headphone from another. Proof positive is not easy to come by, and it's not something to claim lightly. However, I think this test moves us strongly in that direction.
Interestingly in the comments - this is said (by someone who used to be a forum member here - is very conversant with measurements and things like designing EQs etc):
Manufacturing tolerances, my thought exactly.
I really respect what you're doing but that's a big problem in a test where you try to show that burn-in exists. You've seen yourself that different headphone samples (even if they have the same color
and probably even if they're from the same batch) have slightly different channel balance, impedance, sensitivity/efficiency (-> different SPL).
What I'm trying to say is that I think you could achieve the same test results with two brand new Q701s or K701s.
Tyll acknowledges this - which is partly the reason he continued to do the test in the first link which does not have the same flaws (4 months later). You'll note that after that test his conclusions are very different. What has happened in 4 months? Maybe a better and more accurate means of testing break-in on a single pair of headphones.
This is why I ignored the subjective results from his earlier test. It was flawed. All he did was tell two headphones apart. Check the AKG threads, you'll find a heap of information there about manufacturers tolerances - and OOTB AKGs sounding and measuring slightly differently.
Sorry - this got a bit long. Next one coming soon.