Prosody plays a vital role in communication, but is one of the most widely neglected topics in language documentation. This omission is doubly detrimental since intonation is unrecoverable from transcribed texts, the most prevalent data sources for many indigenous languages. One of the underlying reasons for the dearth of prosodic data is methodological. Modern technology has removed technical barriers to recording the appropriate data, but traditional methods of elicitation still inhibit accurate documentation of linguistic structures at or above the phrasal level. In addition, these methods do not facilitate the mobilization of linguistic documentation. In this paper, we present techniques that we have developed that address both these concerns: 1) eliciting prosodic data for theoretical analysis, and 2) producing linguistic materials that can be useful for educators and curriculum developers. Highlighting advantages and disadvantages, we compare traditional elicitation and text-gathering methods with two non-traditional methodologies using non-verbal stimuli. These two non-traditional methodologies are aimed at collecting: 1) spontaneous conversation (either unguided, or task-oriented), and 2) partly scripted conversation (aided by multimedia tools). The methodologies are illustrated with original fieldwork on focus and intonation in two related, endangered Interior Salish languages--Nlhe7kepmxcín (Thompson) and St'át'imcets (Lillooet).