We examine the (potentially nonlinear) relationship between inequality and growth using a method which does not require an a priori assumption on the underlying functional form. This approach reveals a plateau completely missed by commonly used (nonlinear) parametric approaches - the economy first expands rapidly with a large decline in inequality, plateaus when inequality remains relatively stable, and then decreases rapidly with a large increase in inequality. This novel finding helps reconcile the conflicting results in the literature (using different parametric assumptions and datasets) and has important policy implications.