University of Tokyo
English conversation exhibits sets of items such as uh-huh, uh-hn, uh-mm, uh-uhmm, un-hn, uh-uh, and uhh-uhh, which, although clearly related, show great phonetic variation. This paper explains this as due to the presence of sound symbolism: specifically, each of the 9 sound components common in such non-lexical conversational utterances (grunts) bears a meaning or function which is fairly constant across phonetic contexts (for example, the meaning contribution of /m/ in um, in myeah, in mm-hm and so on) and across pragmatic contexts (for example, the meaning contribution of /m/ in fillers, back-channels, interjections, sentence-final particles and so on). For example, /m/ indicates that the speaker considers the conversation topic to be deep or significant, giving it extra thought; nasalization indicates that the speaker considers the conversation to be covering something that is already shared knowledge; breathiness and /h/ indicate that the speaker is actively engaged in the conversation, creaky voice indicates that the speaker is somewhat withdrawn from the conversation, and reduplication indicates that the speaker lacks anything to say. Evidence for these sound-meaning correspondences is found in minimal pairs, and in statistical analysis of all 316 conversational grunts occurring in a small corpus of American English conversations.