We have observed the longitudinal NMR in 3 He-£down to 0.3 T c at 0.4, 10.0, 20.0, and 29o3 bars. From the measured ringing decay time T Rt we deduce the strong-coupling correction to the gap parameter and the quasiparticle-spin scattering cross section.We have performed direct observations of the phenomenon of longitudinal nuclear magnetic resonance in superfluid 3 He, predicted by Leggett, 1 at temperatures far below the superfluid transition temperature T c . Previous studies have been made close to T c in adiabatic compression cells 2 " 4 by sweeping the pressure and in magnetization ringing measurements. 5 The experiments reported here have been conducted on 3 He-J3 at the pressures of 0.4, 10.0, 20.0, and 29.3 bars down to T ~ 0.3T C . The oscillatory response of the longitudinal magnetization following a short sinusoidal magnetic excitation is recorded directly as a function of time, providing measurements of the ringing relaxation time T R . These experiments, already briefly accounted for, 6 have been extended and analyzed in the framework of kinetic theory which is exact in the lowtemperature, collisionless limit except possibly for those strong-coupling corrections which cannot be reduced to a renormalization of the BCS gap. This analysis yields the values of the gap parameter A at T = 0 and of the collision probability of two quasiparticles in the superfluid.Let us first describe the experiment,, The sample chamber is attached to a copper nuclear demagnetization stage capable of cooling 3 He down to \T C for periods of time of several days with a negligible warmup rate c The NMR coil (360 turns of 0.07-mm Ag wire wound by sections on a diameter of 9 mm, over a length of 15 mm) is connected to a fast-recovery (<20 jtxs), low-noise preamplifier and excited by a transmitter delivering up to 50 V (peak to peak). One of the problems of experimental work on the an-isotropic superfluid is the characterization of its orientational state, i 0 e., in the case of 3 He-5, of the axis of rotation n of the order parameter. In our cell the vector n is constrained to a uniform texture by a stack of parallel circular platelets, made of 0.1-mm-thick Mylar. The axis of the stack is parallel to that of the NMR coil. The space between adjacent disks is 0.4 mm wide and 8 mm in diameter. A 1.2-mm hole is drilled at the center of each disk. A dc magnetic field H 09 which ranges from 0 to 400 Oe, is applied in a direction parallel to the axis of the coil, and hence to the direction of n, and does not compete with the orientational effect of the walls. Its main effect on the magnetization ringing is to bring to sight the slight residual misalignment of the vector n (less than 3° across the sample). This comes about because the longitudinal frequency, which is Oj|(T,P) in zero field in the absence of n defects, becomes £l\\(T,P)\cos(n,H 0 )\ in high fields. The best-quality data were obtained in zero field, the high-field results being somewhat less reproducible. The ringing is excited by an rf field H 19 generated in the pic...