In this paper we present an analysis of chemistry texts (mainly textbooks) published during the first half of the 20th century. We show the evolution of the explanations therein in terms of atoms and of atomic structure, when scientists were interpreting phenomena as evidence of the discontinuous, corpuscular structure of matter. In this process of evidence construction, new contributions from physicists and physical chemists that were incorporated to chemical research acquired 'chemical' meaning, since they were related to research questions that genuinely came from chemistry. Conversely, the core ideas of 19th-century chemical atomism, among which we must highlight valence and Mendeleev's periodic system, provided 'clues' for imagining an atom in terms of the elements adjusted to their chemical behaviour, which changed periodically as a function of atomic mass. With this, chemistry ceased to be a descriptive science and began to be a 'law-based', theoretical science. Little by little, chemistry teaching became the teaching of the internal structure of atoms, which were arranged in the Periodic Table according to criteria and 'construction rules' related to quantum mechanics. We pose the question: 'how can we now teach general chemistry in a way that does not disregard current knowledge about the structure of the atom yet, at the same time, gives priority to chemical criteria, thus making such structure useful to interpret chemical change?'. © 2008 Springer Science+Business Media B.V.