Identification of low-order linear multiinput/multiouput models can lead to accurate descriptions of the dynamic behavior of a continuous crystallization process. While open-loop experiments exhibit an oscillating crystal size distribution, improved experimental conditions can be established through stabilization of the process with a simple single-loop feedback controller. The resulting closed-loop identification problem is studied using low-order linear multivariable input–output models. Two closed-loop identification methods are applied, one of which was recently introduced to provide accurate approximate models in general closed-loop process configurations. Identification and validation data are obtained from an evaporative pilot crystallizer, and the identified models are validated in terms of time- and frequency-domain responses. A fourth-order, three-input three-output model is shown to describe accurately the process dynamics. The results are compared with a linearized and reduced first-principles model.