This experiment starts with the assumption that the struc¬ture and representation of scientific information should correspond to the cognitive structure assumed to exist in both user and producer of information. The model of in¬vestigation of cognitive representation is based on overt manifestations of concepts and conceptual relations as they emerge in the abstract language of titles of scienti¬fic documents. These are based on a randomly collected sample of forty social science researchers in Sweden. On the basis of this language structure an algorithm has been developed and tested using the cue function of prepositions for automatic conceptual coding. The relevance of the concepts is judged with respect to a schema model containing some basic components of research itself. The algorithm generates the assumed scientific concepts and assigns them to different data registers.