Thank you, Mary, much appreciated! You probably are dealing with more diverse situations than I am picturing. There may be edge cases, where is it a bit hard to decide whether it is more this or more that. I was assuming that most situations are relatively clear, and that trying to cover every eventuality could be more confusing than helping. For HumanObservation, I would anchor my decision around the question whether a human observer (if not by name, then by role/function - and ideally a date or at least a time period) can or could be identified for the individual records: an observer id in a citizen science app, names in a field book transcript, observer names or ids on survey checklists, the reference to “an elder”, this kind of thing. Literature data is typically aggregated at a higher level, and would not be traceable to individual observers.