May I offer a perspective from my experience with linguistics and anthropology? One of my instructors studies a language which has data source as one of its requirements (not unlike how English requires us to use singular/plural). [She told us that her students who were native speakers of this language were much better at quoting their sources than the native English speakers.] Sometimes when people hear that languages with data source exist, they say: "But that means you can't lie in that language!" --- which is ridiculous, because speakers can use the wrong forms deliberately, just as we can say things that we know not to be true in English. An assertion of my marriage: "Jan got married" could be reported with any of five different grammatical forms, which would express the same quality of source information as the English examples that follow: 1. Things directly experienced by the speaker: "I was at the wedding." (I saw them get married.) 2. Things you were told by someone else: "Someone said Jan got married." 3. Things you deduce or infer: "She's calling herself [some naming convention used by married people] so I guess they got married." 4. Non-involvement: "I wasn't at the wedding and I don't care anyway." 5. Historical: "It was so long ago that no one who was there is alive anymore." (No one alive could know.) When I'm weighing a piece of evidence, I find it useful to look at the data in the same way. If you are a contemporary of the bride and groom, the date and place of marriage is direct evidence, given by the officiant, and witnessed by the witnesses. The information about the parents is "someone else said so" information because the bride and groom only know their own parentage from what they were told by people older than they are. If I am reading the certificate now, and telling you the details, and all the participants are deceased, the event is "no one alive could know". If we only have the marriage registration and not the certificate, we can use it to estimate a date range and a place range. It is indirect evidence. Non-involvement might apply if I advise someone not to use a particular online index because I think the index is faulty and the search results are suspect, e.g. some batches of Massachusetts marriage records on Family Search which, according to their "known issues" pages in the Research Wiki, have death dates reported as marriage dates. With death certificates in the USA, we usually have the name of the informant on the certificate. Assume the deceased's spouse is the informant -- they were not present at the birth of the deceased, nor at the birth of either of the deceased's parents, so all of that information was told to them by someone else, and if the deceased is old enough, it is also "no one alive could know" information -- there is no one the informant could consult with who would have direct knowledge. I find it useful to think of things from this perspective, to bring myself back to the question of how do I know something? It's easy to get tangled up in the labeling of primary evidence / secondary evidence and lose sight of what is really going on. And in my mind, it is more important not to lose sight of whether the source is original or derivative as in this example from Elizabeth Shown Mills about examining church records, posted on 14 January 2014: https://www.evidenceexplained.com/quicktips /document-day-analyzing-church-records In that case the marriage record was a derivative source and was compiled from two different records. Yes, you can argue about what bits are primary and secondary, but if you aren't certain that the two compiled records actually belong to the same person or not, what good is that? Jan