> I don't think there can be such a clear distinction. A birth > certificate is > just a copy of part of a page in a register book. A census > household is > equally just a copy of part of a page of a census book. So if > you look at > it like that there is very little difference between them. It > is all down > to personal preference as to how you view it. I agree that the decision is largely down to personal preference, but not that the two examples are equivalent. There are two important things when using sources: - can I find where I got this data from - can others follow me and retrieve the same data? A certificate is indeed merely, in theory at least, a transcript from the register and often, as most certificates these days are from the GRO, a transcript of a transcript. However, a certificate is, of itself, a legal and standalone document and therefore deserves to be a source. More importantly, it is possible - and I have examples in my tree - where the same event will result in different certificates because different registers were used to produce the certificates, or because of transcription errors in the copy register, or simply because a handwritten certificate was mis-transcribed even though all registers for the event contain the same data, or because of birth re-registration. Thus it is critical that, when citing a certificate, I make it clear that the certificate I used is the one in my possession (the repository being my house) rather than a certificate which someone else might obtain and which may have different data. A single source of "Birth certificates" would not satisfy the second objective above. Conversely, there is only one census (for each year) and, without using indexes, it's not possible for a researcher following me to find the data I used without knowing the class, piece, foilo and schedule reference. This is why, using 1891 as an example, I use a repository of TNA, source 1891 census class RG12, and citations such as RG12/1234 f56 p78 s9. A single source of "1891 census", appropriately cited, will satisfy the second objective above. Thus neither method 1 nor method 2, used exclusively, can fulfil the requirements: it's necessary to use a mixture of the two dependent on the source. Alan