We already saw that, at level 2, a label-type model activates a ZM model. The next step is to activate not the whole model, but only some associated truth of the ZM-model. In this way, the brain has to manage a reduced quantity of information and so becomes more efficient. This is a critical point, because it is the barrier to separate the animal world of human world. Thus, there is a ZM-model and an associated label-model. The problem is to associate to the label-model only some truths generated by the associated ZM- model. A ZM-model is an image model, and so its truths are also of image-type. The problem is to record such truths in a different way, based on a totally new function. MDT cannot indicate how exactly this facility works. The theory is not concerned with the technological implementation of the functions. The theory just says that some truths generated by a ZM-model have to be recorded in a different way. In this way, the label-models become words, and the associated truths become symbolic definitions of the words. On level 3 a label-type model can activate an associated ZM model, but it can activate only a collection of truths as well, which are different from the 'ordinary' image-truths of the ZM. It is possible that the General Communication Language (GCL) appeared based on this facility. The presence of a GCL in a brain will characterize that brain as a human brain. Example: when the word "dog" is heard, it is very probable that we activate at least one suitable ZM. But when we use the sentence "I go to the forest with a gun and a dog", it is very probable that we do not activate any ZM-model. The sentence is understood based on symbolic models and based on logic and so we do not need any image-model. In this way, the quantity of information that has to be processed by the brain is reduced very much. The image models will be used only when we have to make a precise model of the action. The human brain continues to evolve with level 4. On this level we have words and associated symbolic definitions, but no ZM-image-model. Example: Let's take the following words: "this apple", "apple", "fruit", "food". "This apple" is associated with a pure image model. "Apple" is a concept type image model. "Fruit" and "food" cannot be associated with any image model (we cannot imagine what is fruit or what is food). So, on level 4, the human brain can make and operate symbolic models without any connection with image-models. On this level it is possible to develop logical and mathematical languages and, in this way, to make positive sciences associated to the external reality. Example: Newton's Mechanics is a symbolic model associated with the physical bodies. The basic terms of this symbolic model are mass, space and time. None of these terms can be associated with image models.