“Approximate truth” refers to the principle that border cases should be analyzed by solving generic cases and solving border cases as limits of generic ones (Brennan et al., 2008). Our study experimentally explores whether this conceptual principle is also behaviorally appealing. To do so, we focus on perfectness (Selten, 1975) and use his example game with (no) multiplicity of (perfect) equilibria. Distinguishing three uniform perturbation levels, we check for monotonicity (all players react monotonically to the perturbation level) and then explore the behavioral relevance of “approximate truth.”