Abstract. Identifying how firms compete, that is, their conduct, is central to empirical industrial organization and antitrust policy. The standard approach tests candidate conduct models using markup-shifting instruments. When such instruments are weak or unavailable, existing tests have limited power to distinguish among competing conduct hypotheses. This paper develops an instrument-free procedure. In place of an instrument, I use economically motivated bounds on marginal cost. Under the null hypothesis, the marginal cost implied by a candidate conduct model, evaluated at the true demand parameter, must lie within these bounds. This turns conduct testing into a system of linear inequality restrictions on the implied marginal costs. I adapt the Conditional Chi-square testing approach of Cox and Shi (2023) to this framework and establish asymptotic validity of the resulting test. The test adapts its critical value to the geometry of the binding constraints. It requires neither tuning parameters nor simulation, and its critical value comes directly from the chi-square distribution. I design a structural mixed-logit Monte Carlo simulation to Döpper et al. (2025) to verify the test's finite-sample performance. The test controls size well under two common demand estimators, GMM via PyBLP and the Conformant Likelihood Estimator with Exogeneity Restrictions (CLEER) via GruMPS.