In 1924, Einstein predicted the occurrence of condensation in an ideal gas. This seemingly self-contradictory prediction derived from his application of Bose’s statistical count of light quanta to an assembly of material particles. It appeared in print at the beginning of 1925, and was confirmed by the experimental production of the phenomenon, now known as Bose-Einstein condensation (BEC), in ultra-cold dilute gases in 1995. Although much happened in physics in the intervening seventy years, both theoretically and experimentally, physicists continue to describe matter as composed of elementary “building blocks”, or particles, which they see as the modern version of atoms. At a deeper level, however, the conception of atoms underwent a radical transformation. Explaining the aggregation of particles into complex systems—from the nuclear atom of Bohr’s theory, to molecules, and to macroscopic solids and fluids—demanded an unsettling revision of the atomic model of matter that had enabled the adoption of statistical methods in the first place. In the words of the experimental discoverers of BEC, “atoms have lost their identity”. The loss of classical identity is conceptualized as a result of the so-called wave-particle duality of quantum mechanics, and is manifested most strikingly in a group of laboratory phenomena, including BEC, which are named “macroscopic quantum phenomena”. In this paper, I examine the early steps in the uneasy adaptation of the atomic model to the new understanding of the statistical properties of material particles, from the integration of Einstein’s quantum statistics with the quantum mechanics of multi-particle systems to some key developments that led Fritz London to formulate the concept of macroscopic quantum phenomena.