This paper discusses how to take into account en-route choices in utility-based route choice modelling. A path observation through a network is divided into a sequence of route choices and decision making choices at decision nodes. A driver's dynamic taste for different routes at different decision nodes is described by a mixed-logit specification. Four estimation methods are proposed. Two numerical experiments are designed to illustrate estimation issues arising with the proposed model, as well as to illustrate the potential biases introduced by over-simplifying assumptions on en-route choices. The numerical results show that maximum likelihood estimation based on path observations is unable to recover the true parameters, while the other three methods give consistent estimates. The model fit is significant improved by considering dynamic taste for different routes. Path share predictions diverge greatly from those of the proposed model if the assumptions about en-route choices are simplified. The experiment on a real scale network shows that it is feasible to apply the proposed model in practice.