Your sample is too small to answer that.
Take the brutal fit of (18,12,8) by the exponential that passes by 18 and 8, then its expected value at the middle would be 12, the geometrical mean since the levels are equally spaced. So a Boltzmann fits.
But a (non-natural) linear distribution passing by 18 and 8 would give 13 as the expected middle value, which is very close to the observed 12. A rough estimate of the standard deviation would be sqrt(13), and the observed value would be at 0.3*sigma, excellent as well.
Even a flat distribution of 13 particles per level with SD~3.6 would be only 1.4*sigma away at both extreme levels, still good.
More complicated cases would be seriously difficult to answer, requiring harder math tests like chi-squared to give a mathematical opinion - which, in statistics, would be only one element for your human decision.