In this paper we investigate the effects of herding on asset price dynamics during continuous trading. We focus on the role of interaction among traders, and we investigate the dynamics emerging when we allow for a tendency to mimic the actions of other investors, that is, to engage in herd behavior. The model, built as a mean field in a binary setting (buy/sell decisions of a risky asset), is expressed by a three-dimensional discrete dynamical system describing the evolution of the asset price, its expected price, and its excess demand. We show that such dynamical system can be reduced to a unidirectionally coupled system. In line with the rational herd behavior literature [Bikhchandani, S., Sharma, S. (2000), Herd Behavior in Financial Markets: A Review. Working paper, IMF, WP/00/48], situations of multistability are observed, characterized by strong path dependence; that is, the dynamics of the system are strongly influenced by historical accidents. We describe the different kinds of dynamic behavior observed, and we characterize the bifurcations that mark the transitions between qualitatively different time evolutions. Some situations give rise to high sensitivity with respect to small changes of the parameters and/or initial conditions, including the possibility of invest or reject cascades (i.e., sudden uncontrolled increases or crashes of the prices).