Originally Posted by markus767
So you think the ABX methodolog isn't suitable for determining if an audible difference exists between two signals?
No, this may surprise you but I support DBT... I think blind tests are the only way to settle the issue. However my concern is about the conditions of the tests as commonly done, and the interpretation of the results. Listener training is important but I almost never see any guidelines for this. I mean, what works with one person won't necessarily work with another. So you need to formulate a system that will provide training that maximises hearing sensitivity for each person. How does the listener control their attention? For example, how does the tester give instructions so the listener uses their attention in a consistent
way? If everyone is different, what would be the best way of handling this for each person?
How are the musical selections handled? Are the same musical selections played repeatedly? What about fatigue setting in? How does one factor that into the test results? If after a while sensitivity to any difference is close to zero then the probability of experiencing a null result is going to be incredibly high because you can't reliably discern differences at this stage. How many samples should one do? 20 samples per listener to be statistically relevant? 50 samples? Over a wide range of conditions, in different environments or just one environment? How do you factor in the emotional state of the listener possibly affecting the results negatively? Do you rerun the test? Our perception changes from moment to moment, so one could easily argue that failing a DBT on one day could very well result in a pass the next day.
This is a heck of a time consuming process. Just going in and doing a DBT and getting a null result doesn't necessarily paint the full picture here. One could argue that a null result is a reflection of poor implementation of the controls, much to the dismay of the proponents, but I'm not interested in biased, prejudiced answers. Independent verification of the results would be a great idea and I never see any ... because I don't think any exist. Just taking the results at face value means little to me. As it should mean little to anyone else.