We present experimental evidence for nuclear spin selection rules in chemical reactions that have been theoretically anticipated by Quack [M. Quack, Mol. Phys. 34, 477 (1997)]. The abundance ratio of ortho-H-3(+) (I = 3/2) and para-H-3(+) (I = 1/2), R = [o-H-3(+)]/[p-H-3(+)], has been measured from relative intensities of their infrared spectral lines in hydrogen plasmas using para-H-2 and normal-H-2 (75% o-H-2 and 25% p-H-2). The observed clear differences in the value of R between the p-H-2 and n-H-2 plasmas demonstrate the spin memory of protons even after ion-neutral reactions, and thus the existence of selection rules for spin modifications. Both positive column discharges and hollow cathode discharges have been used to demonstrate the effect. Experiments using pulsed plasmas have been conducted in the hollow cathode to minimize the uncertainty due to long-term conversion between p-H-2 and o-H-2 and to study the time dependence of the o-H-3(+) to p-H-3(+) ratio. The observed R(t) has been analyzed using simultaneous rate equations assuming the nuclear spin branching ratios calculated from Quack's theory. In p-H-2 plasmas, the electron impact ionization followed by the ion-neutral reaction H-2(+) + H-2--> H-3(+) + H produces pure p-H-3(+), but the subsequent reaction between p-H-3(+) and p-H-2 scrambles protons. While the proton hop reaction (rate constant k(H)) maintains the purity of p-H-3(+), the hydrogen exchange reaction (rate constant k(E)) produces o-H-3(+) and acts as the gateway for nuclear spin conversion. The value of R(t), therefore, depends critically on the ratio of their reaction rates alpha = k(H)/k(E). From observed values of R(t), the ratio has been determined to be alpha = 2.4. This is in approximate agreement with the value reported by Gerlich using isotopic species.