Abstract
After the error analysis of four SLI children (mean age 7.2 years) and after facing the problem of mapping their language difficulties in a model of language production, it is obvious that the question about what knowledge and operations are impaired has to be restated. Our evidence shows that there is hardly one specific linguistic pattern. We therefore suggest two lines of analysis: 1. The inadequacy of the pre-verbal message preparation as a cognitive problem and 2. The context dependence of rule governed linguistic formulations. These two lines of research are only outlined, but are worth trying and can be a step forward into the comprehension of the core problems of the SLI children.