This paper deals with what is called here paralinguistic organizers i.e. textual devices like underlining, brackets, footnotes, and, on screen, pop up fields,... In the frame of the theory of discourse comprehension (van Dijk & Kintsch, 1983) it is proposed to view technical texts as tools, more or less efficient, that readers use to build their representation (situation model). It is known that the schematic superstructure of a text can significantly enhance its comprehension and memorization. It is assumed here that another kind of text structure could play a similar role i.e. the structure based on the main points / secondary points distinction. More generally a text can be viewed as a set of textual units, each one corresponding to a communicative intention of the writer. In other terms, each textual unit takes in a propositionnal content and an illocutory value. The distinction of secondary points from the main points corresponds to a value of “importance minoration”. Among secondary points, a number of more precise illocutory values can be found in technical texts such as commenting, illustrating, guiding the reading process, synthesizing... Our conjecture is that a clear distinction by the reader of this illocutory values should contribute to text comprehension and memorization. Analyzing a corpus of various technical texts we built a typology of textual units on the basis of the writer' intentions. In order to test the cognitive reality of such a typology, subjects were requested to analyze texts into textual units and to categorize these units according to the typology, on the only basis of language (the paralinguistic organizers were removed). The results showed that, except for subjects trained to text analysis (students in psycholinguistics), such a categorization was a difficult task. Paralinguistics organizers such as parenthesis, footnotes, and pop-up fields on screen, should play an important role signaling the writer's “minoration” intention when secondary points are concerned. In two experiments the effect of such paralinguistic organizers was studied. Groups of subjects were requested to read a same text with and without organizers. The subjects had to read texts “from beginning to end” knowing that they would have to answer questions on the text. The results did not confirm the common signification (minoration) of the studied organizers. Putting textual units in brackets or in footnotes did not decrease significantly their memorization ; putting textuals units in pop up fieds increased their memorization. A last experiment again investigated the role of pop up fields compared to brackets but in an information retrieval task (in a small basis of texts) instead of a “from beginning to end” reading task. The results showed that, in this kind of task, putting secondary pieces of information into pop up fields significantly speeded the search process compared to a condition in which the same information was displayed in brackets. In conclusion it is argued that selecting information is an important subprocess of text processing, even in “from beginning to end” reading task. In parallel with the increasing production of electronic documents, researches on typology of textuals units and on type-signaling devices are needed toward efficient “access structures” for technical texts.