How do we make decisions or process information about the world?
Do we perceive the world directly? Or do we use some past knowledge of the world to inform our decisions or perceptions?
To investigate time and many other things, the Bayesian framework is a simple and elegant way of exploring such questions. Simply, prior knowledge is combined with incoming sensory information in order to generate a posterior distribution of which an estimate of some property is taken.
Let's apply this to the perception of the time between two beeps.
If, over the course of an experiment you on average experience intervals that are say, 1000ms long (blue dotted line below), then what happens if you then experience an interval between two beeps that is say 600ms (red line) ? Do you think it was indeed 200ms, or does your past experience bias your perception?
If we combine the past experience with the present sensory information, then we get an estimate that is somewhere in between (c. 800ms green line). This is an extreme example but you get the picture.
In the Bayesian framework, things that you have just experienced look more like things you have recently experienced.