Anything that affects frequency response and that is audible by virtue of the effect on frequency response will manifest in measurements of frequency response. This is tautologically true. Whether you are able to identify the anomaly by visual inspection of the response curve is a somewhat different question. But if the anomaly, whatever its nature, audibly affects the frequency response, there will absolutely be evidence of it in the measured frequency response.
As for how you know whether a spike observed in the frequency response is caused by a cabinet resonance vs. something else, this is not easy in the case of panel resonances, because the affected frequencies are inherently difficult to predict. But the other kind of enclosure resonance, i.e., standing waves that set up inside the enclosure, are straightforward to predict given the interior dimensions. If you observe a pronounced spike at some specific frequency and it is so strong as to suggest a resonance of some sort, and if the corresponding wavelength is one of the wavelengths for which standing waves are predicted in accordance with the interior dimensions, there is then very high likelihood that the observed response spike is associated with the standing wave. ... .
Edit: The particular algorithmic formula, by which the wavelengths and frequencies for standing waves that set up within an enclosure may be calculated, was not so very pertinent to the points I was making. Nevertheless it has been brought to my attention that I screwed it up, in using the 1/4 wavelength, odd integer rule. The essential fact is that a pressure antinode must exist at both hard reflective surfaces, in order for a standing wave of a particular wavelength to set up between the hard reflective surfaces. But it doesn't really matter whether it is a node or an antinode, because it must surely be the same for both hard surfaces, i.e., either a node at both surfaces or an antinode at both surfaces, and regardless of which of the two happens to be correct, 1/2 of a wavelength must match the distance between the two surfaces (or any integer multiple of 1/2 wavelength). The distance D = n x lambda/2 => lambda = D x 2/n where n is any integer, even or odd. Note, moreover, that the frequencies of "room modes" are a less straightforward calculation. The equation may be found here:
https://en.wikipedia.org/wiki/Acoustic_resonance, a Wikipedia page that provides a good mathematical explanation of resonance. In the simplistic case where all three dimensions are the same (i.e., a cube), the wavelengths (for various integer values n) are reduced by the factor approximately .58 (the reciprocal of the square root of 3), compared to the corresponding wavelengths (corresponding values of n) for a single pair of hard reflective surfaces.
I wish to note that the web page that was referred by ctrl in his comment pointing out my error is a web page (sengpielaudio.com) that I have previously invested some time in and that I disdain. The problem is partly that the English translation from German was apparently done using a bad robot, but this only scratches the surface even if it happens that the translation is the primary reason that it is so difficult to make sense of it. Of course it makes perfect sense to someone who doesn't need the explanation; this is generally how it is with bad explanations. A good amount of effort was invested in the individual graphics but precious little effort was invested in the organization of the page or in the written explanations. There are content boxes with borders where the key piece of information that identifies the scenario to which the information in the box applies is located at the bottom of the box. Inside these boxes are found many run-on equations where the "=" sign occurs multiple times in a single mathematical expression, which is improper. And the explanations are just not good, well-written explanations. Another page on the same site endeavors to argue that sound power is the "cause" and that sound pressure is the "effect". This manner of differentiating cause and effect is wholly contrived; the page where this is done is gibberish. The sengpielaudio.com page says that this Wikipedia page (
http://en.wikipedia.org/wiki/Standing_wave) is "Not good". This Wikipedia page provides a mathematically correct, generalized derivation of standing waves. A key feature of this derivation is the use of the
sum-to-product trigonometric identity to obtain the characteristic waveform equation for standing waves. Through this analysis it is demonstrated mathematically that nodes occur at even multiples of a quarter wavelength and that antinodes occur at odd multiples of a quarter wavelength. (Pressure amplitude is always 0 at nodes, whereas at antinodes
peak pressure is twice the peak amplitude of either the incident or reflected wave alone.)
Both of the Wikipedia pages to which I have given links (
https://en.wikipedia.org/wiki/Acoustic_resonance and
http://en.wikipedia.org/wiki/Standing_wave) are well-written, well-organized, and mathematically meaningful.