Most tube amps have a high output impedance, usually from about 1 to 4 or 5 ohms, whereas most solid state amps have a low output impedance, less than 0.1 ohm. Hence, tube amps do not have as flat a frequency response into most speaker loads, which are not simple resistors.
E. Brad Meyer in
"The Amp/Speaker Interface: Are your loudspeakers turning your amplifier into a tone control?," Stereo Review, June 1991, p. 53-56, showed some measurements of a good solid state amp and a much more expensive tube amp into two different speaker loads. I am not absolutely sure of the history here but I think Audio magazine and Stereophile began testing the frequency responses of amplifiers into a standard simulated speaker load sometime after that. Anyway, the biggest difference in the signal seems to be in the frequency response rather than in distortion.
Here are the measurements of one tube amp that no doubt will sound different into most speaker loads:
http://www.soundstagemagazine.com/measurements/atmasphere_ma1_mkii2/
Check out Chart 1. This amp is pretty flat into a resistor, but the green trace shows its response into the simulated speaker load: definitely not flat. The characteristics of the simulated speaker load are shown in the explanation of the measurements BHK Labs makes for Soundstage :
http://www.soundstagemagazine.com/measurements/test_amplifiers.htm
Now, that amp has an output impedance of 10.5 ohms, so it is somewhat extreme. However, most tube amps do not have a flat response into the simulated speaker load or into the load of most speakers, so of course they are likely to sound different than amps (tube or SS) with a low output impedance.
Most solid state amps have an output impedance of less than 0.1 ohms, so their frequency response into a speaker load is almost as flat as it is into a resistor.
The old Audio magazine included the measurement of an amp into the simulated speaker load and so does Stereophile. You can easily check out the measurements of amps in reviews on the Soundstage and Stereophile sites to see the results for various amps when driving the simulated speaker load.
The difference in frequency response into different speaker loads seems to me to be the biggest difference in the signals coming from amps with low and high output impedances rather than the distortion. Whether you like the results, of course, is up to your personal preference, but there is nothing magic about it.
