We introduce probabilistic (modular) embedding as an extension to the well-known notion of modular embedding conceived to capture the expressiveness of stochastic systems, focussing here on tuple-based probabilistic languages.