I'm not sure what all the fuss is about. Jim Austin's opinion piece seems pretty fair and accurate to me.
The first 6 paragraphs praise and acknowledge the research that has come out of the Harman group (i.e. Toole/Olive). It's not until paragraph 7 that Austin raises the point that while Harman's target measurement characteristics tend to be preferred by most listeners, by no means do they apply to all listeners.
I agree with this. For instance, Olive's listener preference formula isn't nearly as good as people seem to think it is. If you take the time to read the original work by Harman, you'll quickly discover a few things.
As an example, folks often quote Harman's model as having "86% correlation" as if this should end the conversation. But what does that really mean? For starters, the so-called 0.86 is the Pearson r, which means R-squared is actually only 74%. In plain language, Olive's objective measurement model only explains 74% of the variation observed in listener preferences. This is the model that incorporates things like smoothness, response across a listening window, and slope. Here's what that looks like graphically below (it's from Olive's convention paper).
For a given predicted "score," based on objective measurements, look at how much variation there is in actual listener preferences! For a "5", the range is 3-7, which is a HUGE variation in listener preference. And BTW, it requires a lot of math to even arrive at that predicted preference rating - it's not something you can just "eyeball" from a series of FR curves.
So the truth is that conformity to what we all accept to be important loudspeaker characteristics is "good," but not "great" at predicting how much listeners will like it.
View attachment 73538
Moreover, as some of your know, listener preferences for the amount of bass and treble in reproduced music varies by age, gender, years of listening experience, country of origin, AND musical content!
In a Harman experiment, they gave listeners a pair of headphones, the same pre-selected music tracks, and asked them to adjust bass and treble tone controls to their preference. Here are some representative slides illustrating some of these differences. They're from a Harman slide deck available on listeninc.com.
View attachment 73539
View attachment 73540
View attachment 73541
How can there be a precise "target curve" when people of different age, gender, country, experience, etc. have different FR preferences? There can't. The Harman "target curve" represents the result of all of their listeners combined, BUT it doesn't depict the variation that their listeners exhibited. (A better way to present the target curve would be to include vertical error bars, in my opinion.) People look at a precisely drawn line and assume that it is a precise target. Harman's own research indicates that it is NOT.
Bottom line: Objective measurements are only somewhat predictive of listener preferences, and even within those objective measurements, there is variation in individual preferences that are valid (unless anyone wants to argue that the preferences of a 15-25 year old male with 10 years of experience is the "gold standard," and everyone else's preferences are wrong).