Purpose -The purpose of this paper is twofold. The first is to estimate the correlation between market activity and volatility on an exchange that does not use continuous auctions to find prices. The second is to estimate the sensitivity of that relationship to differences in opinions across traders regarding asset value. Design/methodology/approach -Both objectives are accomplished by using seven years of trader-level data from the Tokyo Grain Exchange, which uses rapid sequences of Walrasian tâtonnement auctions to discover prices. On the TGE, only one futures contract trades at any given time and all of a commodity's futures contracts are auctioned in a rapid sequence, with only seconds between a sequence's auctions. The results are interpreted under the hypothesis that this design causes traders' beliefs to become more accurate and more uniform as a sequence progresses. Findings -Intraday volume is u-shaped while intraday volatility is downward sloping. The volumevolatility link is positive and stays constant or strengthens as traders' beliefs about value become more precise. The link is driven by trades originating from small futures commission merchants, especially those trades entered on behalf of customers. Research limitations/implications -Evidence that accounting for cross-correlations when estimating volatility can have an important effect on estimates is presented. Researchers are encouraged to further explore the implications of cross-correlations. Practical implications -The paper includes implications for existing theory, the measurement of volatility, and the design of central exchanges. Originality/value -This paper uses the TGE as a natural laboratory to test theory. It is the first such study to use data from an exchange that does not use continuous auctions, and the first to document the simultaneous existence of u-shape volume and downward-sloping volatility.