What is the proper loudness for listening?

I've found that Paul McGowan has a point: especially with acoustic music (solo, unamplified voice; solo acoustic instruments or small ensembles; piano), the volume usually sounds best when it reproduces accurately the volume that instrument would produce in your environment. Therefore, even a piano can easily be played too loud; if your room is small, a grand piano would overwhelm it, and not just by taking up too much space. One of the things a good audio system should do is to accurately recreate the size of the instruments being played. Perceived size is not entirely a matter of SPL, but they are correlated.

When it comes to amplified music (rock, most jazz, electronic, etc.), this principle is harder to apply. As long as your amp isn't driven into clipping and the high frequencies still sound sweet, your pain threshold is about right for a band like Tool or Massive Attack. After all, that's where the volume would be in a live show.

Be aware, by the way, that "weighting" on SPL meters is crucial. Most are "A" weighted (note that the OSHA standard ditusa cites is given as "dBA"). Most laptop SPL meters I've seen, and most inexpensive hand-held meters on eBay or Amazon, are ONLY "A" weighted. What this means is that frequencies below 100 Hz are NOT registered at all! So if you're listening at, say, 70 dB with an "A" weighted SPL meter, you're probably hearing music that would measure 80-90 dB in "C" weighting, or even more. I learned this the hard way.

Thank you, @ditusa. These are very informative links.

Note that, according to your third link, "The A-weighted sound level discriminates against low frequencies.... In this setting, the meter primarily measures in the 500 to 10,000 Hz range. It is the weighting scale most commonly used for OSHA and DEQ regulatory measurements. The C-weighted sound level does not discriminate against low frequencies and measures uniformly over the frequency range of 30 to 10,000 Hz." This is just what I maintained in my earlier post, except that A-weighting actually cuts off at 500 Hz, rather than 100 Hz—which, of course, makes my point even more important.

However, I will admit to some puzzlement about the clause I elided: that A-weighting "discriminates against low frequencies, in a manner similar to the human ear." In my experience, audio system measurements are almost always given with C-weighting. Sometimes this is explicit, but even when the weighting filter is not specified, the numbers seem to be C-weighted. After all, we do certainly prize systems, and speakers, that produce sound below 500 Hz! To eliminate these sounds from the meter’s measurement because the human ear is less sensitive to them would not be what an audiophile would want to do. For one thing, low frequencies are felt as well as heard. Surely you (ditusa) are not advocating for measurements that ignore low—and high, for that matter—frequencies!

In any case, when I see the dB levels folks on this forum cite as preferred listening levels, and then try to match them on my system, I find that the higher levels (90 and above) MUST be C-weighted. My ears begin to bleed when I approach 90 dBA, but 90 dBC is just excitingly loud.