In this work we create a preliminary model of dependency between person's capabilities and statistical features extracted from transcribed monologues. People of different age and educational level were asked to paraphrase the same short TV-reports and afterwards to pass a verbal part of an intelligence test. Since such kind of data is very time consuming, we created just a preliminary model on an insignificant amount of data to be sure that we are going in the right direction. Genetic programming was used for developing an approximation model of person's capabilities on the base of collected data.