Sorry, I didn't mean to disparage your measurements. I was indeed thinking more along the lines of the NRC measurements.
However, that doesn't change my point.
Whilst your measurements of the S400 are good, your measurements of the KH80 are even better.
Likewise the M2 vs the Revels.
If a speaker that measures better doesn't sound better, then what is the point of obsessing over measurements?
Either we are measuring the wrong things, or we accept that preference outweighs reference and dive back into the funhouse of audiophile subjectivity.
I admire your attempts to include measurements in your reviews.
But unless you (and others) can reconcile the disparity between what your graphs show and what your ears prefer, we are a long way from being scientific.
No worries, I never thought you were disparaging my measurements. =]
That said, the point I was trying to get across with my earlier post is that we sometimes
interpret the data incorrectly, or at least, place too much weight on certain aspects, especially with limited anechoic data. In that sense, I agree with you that sometimes we "obsess" over measurements. This does not mean the measurements themselves are not useful or that we're measuring the wrong things, but we're working with limited data.
You have to remember that most discussions on ASR are about state-of-the-art or near state-of-the-art speakers. I'm simply saying that once a speaker is as flat as say, the S400, getting the frequency response a little bit flatter like the KH80 probably isn't going to make a huge difference. Same with the revel and M2.
There are diminishing returns for 'flatness,' and these speakers have probably crossed that threshold. Other factors start to have influence. For example, wide dispersion vs narrow dispersion. And there's still some room for preference in that regard.
However, if we were comparing the S400 to a speaker with godawful measurements, that'd be a different story.
Measurements are still useful predictors of blind test results (which are still the gold standard of audio quality for obvious reasons). Remember that Sean Olive actually developed
a formula for predicting preference among a group of speakers - using complete sets of anechoic data - with remarkably high accuracy. We just don't usually have access to that kind of anechoic data and aren't plugging it into formulas, so instead, we on forums eyeball a few graphs and have at it with our best guesses about performance.
That's fine. The measurements are just getting us somewhere closer to the truth, if not the complete truth. Preference is always going to be a part of audio. That's why we correlate measurements to blind tests in the first place.