I agree that human voice is by far the most important instrument for judging speakers, as it's the instrument we're most in tune with, and the instrument with which we're most sensitive to slight distortions. Where I get confused is when I see that this is often used to justify speakers that measure poorly.
It's true that the male human voice has quite a large range. Everything from
this to
this. Typically though, the best speaker for reproducing human voice is going to be the speaker that is most neutral through the that range. Sure, because mix engineers do get it wrong often, some songs will be vocal forward, and on such songs a vocal range suppressed speaker will be better. Likewise, some songs will be mixed vocal range recessed, and on those songs, speakers with peaking in that range will sound best. On average, though, the "wisdom of the crowd" will take over, and just like Francis Galton observed the with the ox experiment, the mix engineers will converge on a neutral response. The best speaker "on average" will be the speaker that is most neutral through the vocal range, as it's the closest to what the mix engineers "on average" thought sounded best. The Harbeth speaker measured here(which is known for it's vocal clarity) has an on/off axis response that is decreasing throughout the vocal range. Because of this, human voices will have their lower tones and harmonics elevated above the higher tones and harmonics. This means that on average, they won't produce human speech as faithfully as (for example) a Genelec 8351b. Of course, I'm speaking "on average", and some recordings will undoubtedly sound more natural with a bass slanted response.
In summary, I agree that human voice is the most important "instrument". But, in my experience, the speakers that are best at reproducing all instruments also tend to be best at producing the human voice instrument.
@Frank Dernie brought up a good point, though. If you've got a particular subset of music that you listen to that tends to be mixed in a particular manner, you want a speaker that is most accurate for that subset of music. I'm a fairly young guy, so much of the albums I listen to are albums I'm hearing for the first time. I want a speaker that sounds best "on average", but older and more experienced listeners may have more specific needs.
PS: The Harbeth measured here is an old model, so I'm not going to judge an entire brand based on that one review. I'm actually a member of the Harbeth forums, and I'm a huge fan of the brand. I love Alan's honesty when it comes to audio snake oil in the electronics sector. He's an "objectophile"(?
) who often laments on all these amp/dac designers claiming their DACs are "more true to the source". I also love the aesthetics of Harbeth speakers, which is odd, as I generally don't like the vintage look. I plan to own an M40 at some point. I heard it at a conference and was very impressed, and for some weird reason, I love the look.