Determining the last name

[Anders Andersson]

This is in response to a general question in Landskap: Öland: Socknar: Runsten about how to determine someone's last name in the process of research, when there is conflicting information (or no information at all) in the records.
 
It's true that under the patronymic system, simply letting the children inherit their father's last name (as is done by default in some genealogical software) will not work. If you are familiar with the use of patronymics, you could of course assume that all the children born before circa 1850 to parents with patronymic names also adopted their father's first name with son or dotter added as appropriate.
 
However, I'm concerned even this would be too much of an assumption with little factual basis. If a child dies at the age of five, and is never mentioned in the records with an explicit last name, but rather flickan Anna or similar, would it be proper to assume that she even had a last name?
 
I usually leave the last name field blank until I have encountered a record providing an explicit last name for that particular person, simply to play safe and not introduce data pulled from thin air.
 
When a person appears with different last names at different times (say, one patronymic and one adopted surname), I take the adopted surname for the last name in my database, perhaps adding the patronymic as a middle name if it occurs repeatedly in the records. For married women (and in recent decades also men) I retain their maiden names in the database, even if I know that they adopted their spouse's last name as their own upon marriage.
 
Maybe this issue has already been covered elsewhere in some discussion under . However, that list of inquiries about hundreds of individuals doesn't make it easy for me or anybody else to even find out.