One method is to use story-type models. They are not able to keep the control of time, but they are able to record the order of occurence of some information. Even so, this method is not very efficient. A story-type model could fragment. Once it is fragmented, the correlation between the primary information is lost. Note: when a story-type model is fragmented, there is the tendency to reconnect the fragments, based on logic. Many times this reconstruction is wrong, but the impression could be good. The fragmentation of the story-type models can be seen when a person describes a complex situation. During this activity, one could change the order of some facts. Another method, used by the brain to keep track of time, is to use some rhythm-models. Such models are specialized models, which try to guess when something will happen, based on what has already happened before. For instance, if the brain receives a sequence of two sounds, a rhythm model tries to guess when a third sound will occur. The supposition is that such models try to find an algorithm, which will generate the sequence. Such algorithm must be changed on and on, in a fast dynamical way, to predict better and better when the next sound will occur. Such rhythm-models can be used, e.g. to recognize the speech or to understand music. The rhythm-models are not able either to solve the time problem, but they are able to solve some time-related problems associated to fast changing external reality in the field of sounds. Let's analyze a bit this problem. First of all, the rhythm-models are very well developed for human beings, and they are of very low quality for animals. One assumption would be that, compared to animals, the human brain has a very high capacity to make and operate image models, and, due to this, the rhythm- models are so good. But, there are some other facts: the European civilization invented the polyphonic music (the most advanced music). But the European civilization is developing based on symbolic models. It is fair to suppose that symbolic models support the rhythm image-models. We can take into account another idea as well: as MDT considers that the capacity to make and operate symbolic models is generated by a specialized hardware (thus it cannot be produced by a normal evolution process), it is possible that the capacity to make and operate rhythm models was added in the same way. This supposition is supported by the fact that, while some animals are able to make and operate some image models above the level of human beings, their capacity to make rhythm models is unusually low. The problem of the origin of the rhythm models is left open for the moment. Another method to compensate for the time keeping deficiency is to record some pattern-models of the external reality. That is, to record some information based on many M-type models, to build a pattern-model at a specific moment of time, and to recognize the pattern later.