Abstract This paper examines whether trading activity conveys valuable information about changes in market volatility dynamics. We use a modelling framework, in which the market smoothly switches from one state to another, according to the volume level. Results show that large volume drives the high volatility regime for most of the markets, quite consistently with the disagreement-in-beliefs hypothesis. The volume decomposition into normal trading activity and surprising information arrival reveals a reverse threshold linkage for emerging markets. Results support the sequential information arrival hypothesis and highlight the key role of asymmetric information and thin trading in modelling the volume-volatility relationship. The proposed volume-based models provide significant forecast improvements over competing models and offer scope for investors to earn substantial profits.