Convection parameterization is a leading source of uncertainty in global and regional climate models, and a lack of ground truth complicates the assessment of convection scheme performance. Here we test a linear framework for quantifying convective responses, using two models run at convection-permitting resolution, to examine model responses to temperature and humidity perturbations. For the models examined, a grid spacing finer than ∼1 km was required for consistent (hence potentially accurate) responses, implying good representation of convectively-coupled dynamics. The results from the tests at 100 and 250 m grid spacing could reasonably be used in idealized tests of convective schemes, as their spread is mostly small compared to the spread of previously reported scheme behavior.