Sign languages, i.e. language in the visual-gestural modality, are known to make abundant use of grammatical non-manual markers (NMMs) that fulfill functions at all linguistic levels. NMMs constitute a layer on top of the segmental layer, which consists of sequences of locations and movements, and they are capable of spreading over domains of varying size. Their suprasegmental nature as well as their ability to spread suggests a comparison to tones in tone languages, which may also function at the lexical, morphological, and syntactic level. In this paper, I offer a detailed comparison of the behavior of suprasegmentals in sign and spoken languages. I argue that they are functionally equivalent in the two modalities, but that non-manual spreading also displays some modality-specific properties. I tentatively claim that spreading of different types of NMMs targets prosodic domains of varying size: the prosodic word for lexical NMMs, the phonological phrase for morphological NMMs, and the intonational phrase for syntactic NMMs. In addition, I suggest that eye blinks function like boundary tones.