In asynchronous online discussions, users actively reference visual materials (e.g., video, document) to provide supporting evidence and additional context. However, creating and comprehending complex references can be challenging, especially when there are multiple referents to refer, or when a referent is highly specific (e.g., specific sentences in a paper rather than the paper as a whole). To identify users' challenges in making references with multiple and specific referents while using existing discussion tools, we conducted an observational study and a preliminary interview. Based on the design lessons, we built Korero, a discussion interface that aims to facilitate complex referencing actions. For evaluation, we compared Korero against conventional interfaces in two user studies with referencing tasks of different referential difficulty. We found that Korero not only significantly reduces the time and effort in making references with multiple and specific referents, but also shows potential in increasing users' engagement with the discussion and referent materials.