How can we build AI and robots capable of forming internal representations, acquiring language, and participating in symbolic communication alongside humans? Symbol emergence in robotics is a constructive approach that addresses these questions by developing embodied agents that learn language through real-world sensory-motor interaction. Rather than treating intelligence as the manipulation of predefined logical tokens, this framework positions robots as scientific models of human cognition—progressing beyond linguistic, schematic, and simulation models to achieve embodied, operationally closed agents embedded in the world. The chapter also revisits the symbol grounding problem, arguing that the deeper challenge is not grounding formal symbols in perception, but understanding how meaningful symbols emerge, evolve, and become organized through social and embodied interaction—the symbol emergence problem.