I still remember vividly the first research report I ever worked on. My job was to find a David Fitzgerald in East Limerick in Griffith’s Valuation. At the time, back when genealogical dinosaurs roamed the earth, the only way to do this was to identify roughly which parishes had Fitzgerald households in Griffith’s using the old Index of Surnames, and then comb through them all one by one looking for Davids. So I hunkered down and resigned myself to hours of grinding through microfiche after microfiche. But then – what luck! – the second parish I checked, Croom, produced a David Fitzgerald.
As I sat down to write the report, though, doubts started to niggle. Yes, David was a relatively unusual forename at the time, but just how unusual? So I went back and continued looking. And of course, it immediately began to rain David Fitzgeralds: in Ballingarry, Emlygrennan, Ballinlough, Knockainy …
The point of this is not (just) to show how green I was. Without some idea of how to measure the likelihood of what you’ve found, its place on the scale of probability, it’s very hard to interpret it. If you don’t know that half of the population of West Cork is called O’Driscoll, and that Cornelius was extremely common in just area, you’re likely to fall on the first Cornelius [O’]Driscoll you find and install him as your ancestor. The appropriate metaphor is the hoary old one about blind men trying to describe an elephant, each one extrapolating from the particular part they happen to touch.
This is especially important for Irish research before the 1850s. At that point most conclusions are balance-of-probability judgements, not cast-iron certainties.
So is there any systematic way to measure genealogical probability? The short answer is no. Doubt your presumptions and go off to count David Fitzgeralds.
The long answer, as usual, is that it depends. Now that so many records are searchable in flexible ways, you can get a sense from a quick search of just how common name combinations are. So there are 107 Cornelius [O’]Driscolls in the 1901 census, but only five Cornelius Maguires. It’s rough and ready, but do it often enough on different datasets and you’ll begin to develop a genealogical Spidey-sense.
Just don’t trust it all the time.