Metastable NH* radicals in a molecular beam, generated in a discharge, were allowed to collide with target particles (He through Xe rare gas atoms, and H-2, CO, N-2, NO, O-2) in a cell or a crossed jet. Optical emission was observed issuing from the collision zone (and in the case of the jet also from different points along the primary beam). Spectral analysis (similar to 0.13 nm FWHM resolution) revealed two components; (a) a pair of sharp P, R lines ("spikes," originating from the (perturbed) level NH(A (3)Pi, v=2, J=5, F-3, Lambda-component "e"; (b) broad NH(A (3)Pi --> X (3)Sigma(-)) emission in the (0, 0), (1, 1), and (2, 2) bands. Component (a) was shown to be due to a gateway coupling with the (perturbed) level NH(b (1)Sigma(+), v=5, J=5). From the collision gas pressure dependence of the "spike" intensity, relative cross sections were derived. They varied by less than a factor of 3 between He and NO. Weak spike emission was also observed issuing from the NH* beam without collisions. From the exponential decay of this "afterglow" intensity along 20 cm of the beam, the lifetime of the long-lived gateway emission component was found to be 52 mu s, with a beam speed of similar to 1220 m/s (measured using a chopper wheel and a particle multiplier detector). There is also a fast gateway component, having a (calculated) lifetime of similar to 0.21 mu s. It is too close (similar to 1 cm(-1)) to the slow component to be spectrally resolved and is, moreover, much weaker. The calculated branching ratio of the fast and the slow component is 1:247. Experimentally an upper limit of 1:20 was derived from simulations of the observed emission intensity profile downstream from the beam/jet crossing point. It is pointed out that only the weak, fast component of the "spike" intensity should properly be termed "gateway" emission, while the dominant, slow component is better described as being due to an "emission window" at a particular level of the otherwise dark NH(b) state. The broadband component (b) of the NH(A-X) emission is due to direct spin-changing energy transfer from (mainly) NH(a (1)Delta) to NH(A (3)Pi). Surprisingly all target gases except He were effective, although the relative cross sections varied here by a factor of 120 between Ne and NO. NH(a) was identified as the dominant reactant species from the different beam attenuation in the target cell, compared to that of NH(b) (as measured using the spike attenuation). The contours of the intense NH(A-X) bands observed with Xe, O-2, and NO were computer-simulated, yielding high rotational "temperatures" and, with O-2, a striking excess population of the "f" Lambda component (e:f=1:5).