The preterm children have more difficulty minor neurological developmental than the general population even without cerebral palsy. Their oral language difficulties of language and writing have been little studied. The investigators propose to study spoken language has 3 years 1/2 and stimulate or not (depending on randomization) children with phonological weaknesses notemment language. The investigators will evaluate stimulation protocol implied short and precise language in a re-education "say and do" in 20 sessions. A new evaluation of language (BILO and N-EEL), sensorimotor and cognitive constraints (KABC) will be performed at 4 years.

Early intervention on the basis of precise language before closing the window developmental (<5 years) is expected to improve as a result of language and cognitive development of preterm infants.

Very premature babies born before 32 weeks gestation have developmental problems more important than the rest of the population. The study reported 40% Epipage neuromotor difficulties of varying severity. Children with cerebral palsy are down 9% to 6%. However the rate of minor neurological problems and school failure is increasing. Neurocognitive difficulties are many, visuospatial dyspraxia, attention disorders, language impairment. The latter is fairly described in the literature and seems often seen in this population in relation to overall cognitive difficulties. Cognitive assessments of the overall K-ABC of former premature study Epipage are nevertheless often normal. Oral language is fundamental to the written language and learning at school and follows a mature development stages according to specific sensorimotor well known. The investigators hypothesize that minor sensorimotor disorders in the population of premature infants without CP can change structurally and early oral language, including phonology. Accurate assessment of the components of language at 3 years of premature children without CP or neurosensory disorder found (blindness, deafness) has already allowed us to observe weaknesses in the structural components of language. These early failures may correspond to a particular neurological architecture of language observed in functional MRI in the same population of children aged 8 and 12 years (Schaer 2009, Petterson 2002) and may be sensitive to a specific stimulation early before the developmental window architecture of the spoken language is closed (Newport 2002).

Objective: To evaluate the interest of accurate and early stimulation implied short ("say and do") on areas of weakness in the language 3ans1 / 2 of the former preterm infants without CP spotted by BILO.

Method: This is a prevention trial national, multicenter, prospective randomized two arms. In this study, 296 patients with premature frailty of oral language without CP, with or without minor neurological disorder, and without proven pathology of oral language will be included in 5 hospitals and will be randomized to receive or not a specific stimulation (say and do " ). The assessment of oral language by a Computerized Assessment BILO1 (Khomsa 2008) ½ to 3 years will be offered to all parents of children with CP followed without neurosensory disorder or current fact. After obtaining consent, and results BILO, directed by a therapist trained in order to define the criteria for frailty, randomization will be done centrally. Early stimulation precise and short (20 therapy sessions per week) on these areas of weaknesses of language drive will be compared to the absence of specific stimulation. Evaluation at 6 months results will be blinded.

Inclusion is scheduled 24 months for the inclusion of patients with results defining areas of weakness in BILO1dans one of the components of language: an item <10th percentile and / or two items <25th percentile). .

Excluding children with pathological scores BILO1 (<3rd percentile for at least one criterion) will not be included Primary endpoint: difference from baseline and after six months of phonology BILO score (score from 0 to 16 listed) Number of subjects and statistical power: 170 children a year born very preterm live out the CHU of Rouen. One hundred are followed and accept the studies conducted in the service as in other centers who collaborated in the study EPIPAGE (Lancet 2008). According to our preliminary work half the present language problems at 3 years. Thus, 50 children per year would be candidates for the study but 10% are pathological oral language and therefore excluded from the study. 296 children were included in five centers and according to their follow-up rate according to these proportions. Three centers have follow-up of very preterm and 80/an a regional 30/an. According to preliminary data, and assuming a moderate correlation of 0.5 between BILO scores at baseline and after six months, the standard deviation of the primary endpoint was 4.7 units of phonology BILO score. Thus, to test the risk of first kind of b = 5% in bilateral formulation, the total of 296 children included and evaluated (or 148 children per group) to detect with a statistical power of 90%, size effect of 0.379, a difference of 1.8 units average score BILO groupes. Ainsi between the two, the number chosen to detect with a high power low difference between the two groups for the primary .

Statistical analysis: The means of the primary endpoint will be compared between groups by one-tailed t test at the usual 5%. This test will be completed by comparing adjusted based on multiple linear regression model taking into account possible prognostic factors (duration of gestation, bronchopulmonary dysplasia, neurological disorders minors, socio-economic status ...), the center and the initial value ( to 3 years) score phonology. The difference in means between two groups and its confidence interval 95% will be estimated.

Outlook: Creating a cohort followed to analyze the impact of early stimulation specific to the age of 6 and assess our level of prevention of disorders of oral and written language.

Inclusion Criteria:

• Premature < 32 without CP,
• criteria of fragility BILO

Exclusion Criteria:

• blindness,
• deafness,
• abnormal karyotype results in normal or pathological BILO