Agricultural trade amongst and between the United States and China dominates world markets and has been complicated by rapid growth, significant changes in domestic farm policy, intermittent periods of considerable volatility, and, most recently, trade tensions. It is unlikely that a single GARCH process can adequately accommodate this vacillation. Not surprisingly, past literature has shown conflicting results depending on the period considered. We use mixture methods which let the data define the number of possibly heterogeneous volatility regimes. We model the price volatility transmissions for five commodities: soybeans, wheat, corn, sugar, and cotton. Specifically, we estimate, test, and find the presence of multiple regimes using a normal mixture multivariate GARCH model. We identify different economic structures across the regimes. While we find that the U.S. tends to play a leading role over China in terms of spillover effects, when the market state is unstable or highly volatile, we tend to find greater bidirectional volatility spillovers. Most importantly, we show that the standard approach of modelling spillover volatilities as a single regime is not valid.