While several existing datasets can help to address pressing questions on nonviolent resistance, data collection on nonviolent conflict involves several distinct challenges, including (1) conceptual distinctions between the absence of violence, non-violent behavior, and nonviolent direct action; (2) a systematic violence bias in mainstream news reports; and (3) incentives to misrepresent. As a way forward, we advocate (1) collecting data at multiple temporal and purposive units; (2) diversifying source materials; and (3) coding ambiguity as a meaningful substantive variable.