As with biological systems, spoken languages are strikingly robust against perturbations. This paper shows that languages achieve robustness in a way that is highly similar to many biological systems. For example, speech sounds are encoded via multiple acoustically diverse, temporally distributed and functionally redundant cues, characteristics that bear similarities to what biologists call "degeneracy". Speech is furthermore adequately characterized by neutrality, with many different tongue configurations leading to similar acoustic outputs, and different acoustic variants understood as the same by recipients. This highlights the presence of a large neutral network of acoustic neighbors for every speech sound. Such neutrality ensures that a steady backdrop of variation can be maintained without impeding communication, assuring that there is "fodder" for subsequent evolution. Thus, studying linguistic robustness is not only important for understanding how linguistic systems maintain their functioning upon the background of noise, but also for understanding the preconditions for language evolution.