Social interaction is a key feature of our daily lives; humans simply cannot help but interact with one another. This interaction is special with regard to its quantity, but it also shows distinct qualities such as a special propensity to read each other’s intentions. One specific kind of interaction that humans engage in frequently and that exemplifies this particularly well is communication. By producing and interpreting signals in their specific context, interlocutors are able to communicate successfully, even about concepts for which they do not yet share conventional signs. Over repeated interaction, these novel signals can conventionalize, and eventually be culturally transmitted to new individuals. Through repeated episodes of transmission, entire communicative systems, such as languages, can emerge and evolve. In this thesis, I build on the framework above to study how human communicative signals can emerge and become organized via interaction. To this end, I present the results of three empirical studies each concerned with one specific question. The first study represents two artificial language experiments investigating the role of context for the successful emergence of novel communicative conventions, the second study focuses on the evolution of population-level cultural patterns, and the third study aims to relate existing communicative conventions about color terms in natural languages to novel conventions created within a smartphone application. The three studies show the usefulness of combining different methodological approaches – experimental laboratory studies, large-scale online experiments, and massive data sets of online behavior – to address questions at different levels of granularity. Taken together, the studies place individual interactions firmly at the base of both the emergence and organization of communicative signals. As a result of these interactions, entire systems of communication can emerge.