We apply linear and non-linear Granger causality tests to four U.S. price indices and 31 commodity series, which expand a 54-year period (January 1957-December 2011). We find evidence of linear Granger causality mostly from individual commodities to price indices. The latter, however, seem to Granger-cause individual commodity prices in a non-linear fashion. Overall, our estimation results show that Agricultural raw materials (cotton, hides, rubber, and wool), Beverages (coffee), Food (maize, rice, and wheat), Minerals, ores and metals (copper), and Vegetable oilseeds and oils (groundnut oil and soybean oil) display bidirectional linear and non-linear feedback effects vis-à-vis price indices. These findings suggest that not only shocks on commodity demand and supply may impact aggregate price indices, but also that non-commodity shocks, embodied in aggregate price indices, may impact commodity prices linearly and nonlinearly.