IntroductionConsider H,,, the hyperbolic space of dimension n 2 2; the hyperbolic Laplacian A H on smooth compactly supported functions defines an essentially selfadjoint operator on L2( V ) , if dV is the hyperbolic volume. The bottom of the spectrum of -$ A H is known to be i ( n -1)2. Here we suppose that we are given a random distribution of points on H,,, by means of a Poisson point process with intensity vdV, and we suppose that each point is surrounded by a hyperbolic ball of radius p . We are interested in the following type of question:Let A be a small positive number; what is the number of eigenvalues less than Q(n -1)2 + X for the operator -$A, in a ball B , centered around a fixed point 0, of radius N , with Dirichlet boundary condition on JB,, as well as on each of the random balls intersecting B,, when N is large enough? More precisely, we consider the random measure on [i(n -1)2, co), where lBNl is the volume of B,, and Xi, , are the eigenvalues of -$ A H in B, with the above described boundary conditions. In the first section we show that this sequence of random measures almost surely converges vaguely to a deterministic measure m(dX) on [i(n -1)2, co), called the density of states. We study the behavior of this measure near the edge $ ( n -1)2, by means of its Laplace transform /Fe-"'m(dX), for large values of t , and a Tauberian theorem.Such a Laplace transform in the case of d-dimensional Euclidean space and the usual Laplacian, can be expressed in terms of the Wiener sausage WP (neighborhood of radius p) of a Euclidean Brownian bridge by the formula ( m is now a measure of [0, co)):The large time behavior of this quantity was studied in the work of Donsker-