Fred,
IF the speakers are carefully measured in an anechoic chamber and no "smoothing" has been applied to the resulting frequency response curve, then fluctations of 3 dB especially peaks, rather than dips, may be quite audible.
By the way, 1 decibel is defined as the smallest step upwards or downwards in SPL that we can detect, however, such tiny variations are extremely hard or virtually impossible to hear with musical programming. You can, however, hear them with pink noise test signals.
A 3-dB variation is subjectively audible as "slightly louder" (or slightly softer).
In terms of coloration, and audibility, it depends greatly on where the non-linearity occurs. Small variations in response above 10 kHz are essentially inaudible and similar variations in the bass octaves are also insignificant.
In the midrange, where most of the energy of musical instruments is concentrated, such variations are quite audible. The M80 v2s are extremely linear both on-and off-axis through the critical midrange, which explains why they are so accurate and musically transparent.
The technically untutored comments on other boards can be ignored. Some of the most neutral and transparent speakers around include the Axiom M80s, a couple from Revel,and Snell, Energy Veritas, PSB and a few others.
The accumulated data from the National Research Council program in which Axiom participated for many years has shown that it's not only the smoothness of the on-axis frequency response, but the off-axis curves that are equallly important in establishing the overall sound quality in a room--the so-called "family of curves".
Most manufacturers publish curves to which 1/3rd octave smoothing has been applied, which makes them look very nice and smooth for marketing purposes.
The quasi-anechoic room measurements conducted by many manufacturers (and published) are not nearly as precise as speaker curves measured in a large anechoic chamber. Axiom has its own large chamber, a clone of the chamber at the Research Council in Ottawa, which we previously used dating back to the 1980s.
Sometimes correcting a small glitch in the on-axis frequency response doesn't make the speaker sound better, because it will affect the off-axis curves.
Incidentally, be sure and check the M80 v2 anechoic measurements at Soundstage.com. There you'll see the full family of curves.
Hope this helps.
Regards,
Alan
