This article aims to evaluate how teachers, in training and in-service, define the concepts of randomness, probability, chance and variability, fundamental terms in the teaching of statistics. To this end, a printed recording protocol of natural semantic networks was applied to 16 teachers in training and 58 in-service teachers, selected through non-probabilistic sampling. The results provide evidence that the concept with the lowest conceptual density in both groups is variability. Likewise, a greater presence of similar words was observed between randomness, chance and probability, despite being different concepts. Another finding is the association of the inducing concepts to words that are used within the basic lexicon in Chilean Spanish, whose dictionary is not specialized in statistics. It is concluded about the scarce use of technical language by the participants, which would eventually affect the teaching of statistics.