How does natural language evolve? This question dates back to at least Darwin who suggested that words struggle for survival in a process of natural selection. How this natural selection actually works is a mystery. I present an evolutionary view of the lexicon focusing on how words are selected for expressing emerging meanings through time. I argue that central to this evolutionary process is the joint optimization of information and cognitive effort. Formulating this idea in a computational framework of chaining, I show that it accounts for large-scale historical data of language change and traces the origins of word selection in early childhood. I also describe recent work under a similar view on the tradeoff between word selection and creation in the evolution of the lexicon, and discuss future outlook.