Populations of children born very prematurely represent 7% of births, or 15 million babies per year. While death and cerebral palsy rates decrease in Western countries, nearly 40% of these children will present neurodevelopmental disorders. Among their difficulties, school failures are overrepresented when compared to full-term children. Reading acquisition is nonetheless very often achieved. We hypothesized that formerly preterm newborns would present minimal motor disorders which, according to the motor theory of speech perception, would hinder the development of phonology. In a longitudinal work conducted in three parts, we analyzed the reading of formerly very preterm newborns by analyzing their event-related potentials on the EEG. In the first part, within a national research program (PHRC), we included children born very prematurely and tested at 3 years of age to characterize and then stimulate their oral language. We randomized them into a re-educated group (n = 87) and a non-reeducated group (n = 78) (LAMOPRESCO study). The results showed the effectiveness of a short and protocol-specific stimulation program on the development of sensorimotor aspects of language at 4 years of age. Re-educated children showed a trend toward a better progression of their score in phonology and significantly improved scores in the lexicon in language reception (p = 0.02), production and comprehension Phonology scores were correlated with motor, fine motor, and laterality scores. In the second part, we developed a methodology of ERP during reading, with stimuli represented by simple and contrasted words in order to distinguish between the phonological motor way and the visual way. The results in two paradigms in ERP of our normative adult population (n = 19) and normal 12-year-old children (n = 16) concerned the N170 wave. This parameter of choice, which signifies the automatization of reading, allows us to confirm the “tuning” effect and to consider the importance of the stimuli. The recorded population of dyslexic paired children confirmed the existence of maturational fluctuations of the N170 wave and its co-construction with the attentional P100 wave. The positive time course of P1N1 was correlated for all populations of normo-readers and dyslexics at our leximetry parameter. The N170 appears to reflect the specialization and automatization of expert reading, but the P1N1 couple informs us about its neurodevelopmental construction through a probable “top down” effect maturation on the P100. The position of the N170 was positive in 56% of the paired premature infants (n = 18), while it is found in 39% of dyslexic children. In the oddball task, the latency of the P300 target words was correlated with the VL in all studied groups. This complementary result directs us to the importance of visual attention and working memory, especially in children born very prematurely during the specialization of neurons in reading. To achieve our third part, which is the recording of our first-time children, who are now 8 years old, by the cVEVs, we simplified our stimulus words. We recorded new normative populations: normal adult readers (n = 25), matched norma, reading children at 8 years of age (n = 22), and dyslexic children (n = 11). They were all differentiated by the heterogeneous parameters of the N170. In the one-minute reading test, dyslexic children performed less well than children born prematurely (n = 18), themselves different from normal readers. The correct response rate and the reaction time differentiated children born prematurely from dyslexic children. In the oddball paradigm, children born very prematurely but rehabilitated (n = 8) were distinguished from children born prematurely but uneducated (n = 9) at the behavioral lecel by better CRR and longer RTs. Non-rehabilitated children had no oddball effect. We demonstrate the effect of a 4-year reeducation program in short-term oral language that modifies reading strategies at age 8.