Entity state prediction mechanisms are used in order to reduce the number of packets required to maintain a consistent state in a Distributed Interactive Application (DIA). Typically in the case where the entity is representing a participant in a networked game this is achieved by continually comparing the output of a prediction algorithm against a player’s actual state. The state usually comprises position and orientation information in such cases. If the error exceeds a pre-defined threshold value, then an update packet is transmitted, which contains the player’s latest trajectory information. However, obtaining a suitable threshold value remains one of the key challenges that face such entity state prediction techniques. Furthermore, these methods can employ two different threshold metrics. These are spatial, which exploits distance measures, and time space, which uses both time and distance measures. While a spatial threshold value can be arguably determined based on a prior knowledge of the gaming environment, it remains difficult, at best, to obtain a corresponding value for the time space threshold metric. This paper proposes the novel use of user perception as a suitable means to solve the aforementioned problem. Here we employ the most common entity update mechanism, namely dead reckoning, and use perceptual feedback to determine suitable threshold values for both spatial and time space threshold metrics. This involves collecting linguistic feedback on short scenes recorded from a racing game. This technique is compared and contrasted with an alternative method whereby equivalent spatial and time space threshold values are obtained based on a common measure of inconsistency. Details of the experimentation and an analysis of the results are presented within.