In Binary Threshold Public Good (BTPG) games players contribute or not to the production of a public good which is produced if and only if there are "enough" contributors. There is a plethora of equilibria in BTPG games. We experimentally test general theoretical attributes of equilibria and proposals for equilibrium selection. As theory predicts, if the cost/benefit ratio is the same, then subjects play (almost) the same mixture of strategies and, after switching from a positive to a negative frame, the theoretically expected "mirrored" behavior can be observed, i.e. contrary to most linear Public Good experiments we do not find a framing effect. A finite mixture model successfully (i.e. without rejection in a chi-square test) describes behavior in all eight experimental games (same parameters for four thresholds and positive/negative frame). The Harsanyi-Selten theory of equilibrium selection is moderately supported. Efficiency as an equilibrium selection device and also risk dominance are clearly rejected.