Could crowdsourcing be a way to get undone science done? Could grassroots groups enlist volunteers to help make sense of large amounts of otherwise unanalyzed data—an approach that has been gaining popularity among natural scientists? This paper assesses the viability of this technique for creating new knowledge about the local effects of petrochemicals, by examining three recent experiments in crowdsourcing led by non-profits and grassroots groups. These case studies suggest that undertaking a crowdsourcing project requires significant resources, including technological infrastructures that smaller or more informal groups may find it difficult to provide. They also indicate that crowdsourcing will be most successful when the questions of grassroots groups line up fairly well with existing scientific frameworks. The paper concludes that further experimentation in crowdsourcing is warranted, at least in cases where adequate resources and interpretive frameworks are available, and that further investment in technological infrastructures for data analysis is needed.