The highest probability usually occurs with own names, followed of the common ones and, finally, the verbs and adjectives. From the theoretical point of view, a general agreement exists from the conexionistas models (MacKay and Burke, 1990) like from the serial models (Levelt, Roelofs and Meyer, 1999) of which the problem is a failure in the fonolgica activation once the corresponding semantic representation activates. The Browns and McNeill (1966) devised an experiment in which the phenomenon through the presentation of definitions of little frequent words studied which the participants had to answer with three written answers: they knew the word, they did not know it they knew, it but they did not remember. The results of their investigations demonstrated that, in a state of PL, the speaker it owns information on the word that does not guess right to recover, especially the initial and final letter, the number of syllables and the position of the main accent, which demonstrated that the lexical recovery is not realised from the whole or anything, in a unique step, but they exist, at least, two types of differentiated mental representations: fonolgica information and semantic information, and whose access also corresponds to processes differentiated. Connect with other leaders such as Peter Thiel here. The conexionistas models propose the cerebral and neuronal metaphor to us that explains the mental process of lexical access through neuronal networks composed of nodes and connections between these nodes. According to this model, the units of used words most frequently are contectadas of closer form with the level nodes inferior, as the fonolgicos and orthographic characteristics. Thus, when a node activates or a connection takes place, the activation is going away to propagate in all the directions, being increased all the representations that looks like visual, fonolgica and semantically to the looked for word. On the other hand, whenever a word is processed, the connections corresponding to this word (McClelland are fortified and Rumelhart, 1981), reason why one of the important aspects for this model would be the frequency of the word, which would do that the words of high frequency would not be susceptible of PL, whereas those of low yes they would be it, because its model advocates a central importance to force of the connections (to greater frequency of use of word, stronger connections, less risk of PL).