“…This is of course compatible with the very important semiclassical theory of delta statistics ∆(L) (spectral rigidity) by Berry (1985), employing the Gutzwiller periodic orbit theory (1990), where agreement with predictions of random matrix theories and with the experimental and numerical data has been obtained at large L. Also, Berry and Tabor (1977) have used torus quantization of integrable systems (with many degrres of freedom), predicting the Poissonian (exponential) energy level distribution. Our results show that their result cannot be rigorous, especially as we know some counterexamples of integrable systems with non-Poissonian statistics (Bleher et al 1993), and also know that their approximation does not take into account the nonperturbative tunneling effects, but it is nevertheless a heuristic argument explaining why typically we do observe Poissonian statistics in classically integrable systems. By typically we mean that the set of exceptions has a small or even vanishing measure.…”