We present an approach to realize a networked on-line education system among physically separated participants targeted at teaching handwritten characters. In addition to visual and acoustic information, our system provides a haptic channel to intuitively learn remote instructor’s fine motor skills through the sense of touch. The human touch sensation, however, is very sensitive for performance degradation caused by delays, packet losses, and jitters in the networked virtual reality (NVR) systems. Therefore, we show a method to efficiently share virtual space information by compensating for the performance degradation factors. We take a hybrid approach to implement both data consistency by client-server and scalability by peer-to-peer models. The handwritten character learning system is built on the hybrid architecture and realizes the participants to tightly share partners’ haptic motions. We also describe an experiment conducted to validate the effectiveness of the haptic modality for teaching handwritten characters.