1 STATISTICAL STUDY OF THE INSTANTANEOUS VALUES OF THE SOUND PRESSURE LEVEL, GIVEN SOUND EVENTS RELATED TO ENVIRONMENTAL ACOUSTICS. PACS: 43.50. 43.58. 43.20. Alba, Jesús; Ferri, Marcelino; Ramis, Jaime; Martínez, Juan A. Departamento de Física Aplicada; Escuela Politécnica Superior de Gandía; Universidad Politécnica de Valencia Carretera Nazaret-Oliva S/N. Grao de Gandia 46730 (Valencia). España Teléfono (96) 284.93.14 - (96) 284.93.00 Fax: (96) 284.93.09 ABSTRACT In environmental acoustics the knowledge of the time dependency of the sound level provides relevant information about a sound event. In this sense it may be mentioned that the conventional sound level metres have frequently implemented programs to calculate the fractiles (percentiles) of the distribution of instantaneous sound levels; and there are several indexes to evaluate the noise pollution, based on different statistical parameters. For further analysis of sound, and to obtain the commented indexes, it is accepted that this distribution is normal or gaussian. The questions we've tried to solve in this work are the following: First of all, if the time dependent distribution of noise shall be considered as Gaussian in general cases or only in same particular. In the other hand, we have studied how the frequency of the sampling, and the applied exponential time constant (fast slow), affects to the kind of the resulting distribution given a sound event. To get these aims, it has been studied a set of sound events such as the traffic noise in different points of a highly contaminated town, obtaining the conventional "statistics" (moments of n-th order and percentiles), studding the shape of different plots, etc. After the analysis of the measured events, we can advance that in many cases it is not reasonable to consider the distribution as gaussian, but asymmetrical distributions are frequently found. 1.-INTRODUCTION A gaussian distribution verifies: q 1 ) x ( ) x ( x q - =         σ μ - Φ , (1) where ) z ( Φ stands the typified error function (with average 0 and standard deviation 1) , y x q represents some value of x with a probability q (q ) 1 , 0 ( ∈ ) to be surpassed. On the other hand, μ(x) y σ (x) mean respectively the average and standard deviation of the distribution of x. Concretely, for the measure of the statistical distribution of levels from a sound signal, stands: 100 n 1 ) L ( ) L ( L n - =         σ μ - Φ , (2) being Ln the percentile level surpassed during n% of the measurement time. Previous expression can be written in inverse form as: