A clinical thermometer model is proposed, which can be described by discrete Langevin equation. The characteristic of the clinical thermometer is that its reading can only increase with the increasing of its temperature, but cannot fall even when its temperature decreases. "Pause" events are defined according to the non-decreasing character of the reading. Using the random walk thoery, we derive analitically the distribution of the duration of the pause events. Both numerical and analytical results show that the distribution function has the form of a power law D(s)∝s-ξ, indicting the critical behavior in this process.