There's a saying: "Genealogy without records is mythology." It's blunt and maybe it sounds a little smug, but it's true. Truth, or at least as close as one can get to it in this context, is based on evidence. We believe in what the documents tell us. Of course, this is a simplified statement; there are tiers of reliability for documents, as well. Sometimes you'll find records that include contradictory evidence and you're not sure which record to trust, or even why a record might be wrong in the first place. This is where the vitally important critical thinking part of family history research comes in. Let's get technical.
Every time you look at a source, you need to figure out two separate things and use your answers to these questions to drive your analysis of which records to believe and how much to believe them: what kind of source is it? and what kind of information does it contain?
There are two types of records in genealogy: Original records and Derivative records. Other disciplines might refer to them as primary and secondary sources. In order to figure out what kind of record you're looking at, work out What you're looking at. Is it the actual record, or a version of the record that's been reproduced or indexed1 in some way? An original record is a record in its first form that has not previously appeared in another form. For example, a will written by hand on a piece of paper is an original source; if that same will, however, is then reproduced in a book of will abstracts2 (or the transcribed will produced by a suspiciously grinning heir), then it becomes a derivative record, which is defined as a record that relies on other records to exist. Transcriptions and compilations of original records are both examples of derivative records. Or, to put in another way, an original record is the first accounting of an event, while a derivative record would be based on that original record. An additional aspect of an original source is that it was created at or at about the time of the event it documents, while a derivative source will come later.
"Why, yes, I've got Grandpa's will right here..."
Since original records don't rely on any other records to exist, whether through copying or other means, they are the most accurate relator of information. Whenever possible, you want to get as close to the original source as possible, largely because derivative sources can contain mistakes not present in the original. A transcriber, for example, can misread the bride's name on a marriage license as "Tilly" rather than "Milly," and cause you a whole world of trouble, not to mention how confused Tilly is now. Often derivative sources are all we have to work with, and they are valuable tools, but it's important to understand that they have their limitations.
Note: There's some disagreement among genealogists whether a faithful image of an original record is itself an original or derivative record, such as a photocopy or a digital image of the original record. I tend to fall on the side that says that it can be considered an original record, provided that care has been taken to reproduce the record in its entirety.
So! You've looked at your record, and you've come to the conclusion that it's an original record. Let's say it's the enumeration of your grandparents' family in the 1940 United States Federal Census. Since it's an original source -- created at the time of the event, and not relying on any previous sources for its existence -- you ought to be good to go and believe whatever it says, right?
Before you can even begin to think about believing what it says, you need to figure out the type of information it contains. Much as there are two types of records, there are also two types of information: primary information and secondary information. Primary information is those bits of information that come directly from the horse's mouth, as it were, based upon eyewitness information, while secondary information is hearsay ("Well, I wasn't there myself, but Regina swore up and down that Barbara came to church in a PINK dress and scandalized the whole congregation!"3) and contains at least one layer of distance between an eyewitness and the record. As you might expect, primary information is going to be more reliable than secondary information, but figuring out when you're dealing with secondary information can be a bit tricky. Just ask Tilly.
"Wait a second... Where's Milly?"
As an example, Grandma's mother's diary with an entry about the night Grandma was born is primary information, as Great-Grandma was ostensibly an eyewitness to the proceedings; however, Grandma's story that she tells about the night she was born is secondary information, since, unless Grandma was the creepiest baby ever, she doesn't actually remember the night she was born. While she was present, she wasn't an eyewitness, and so her story has either been transmitted through her mother's diary or through what she was told as a child. That's right, boys and girls, I'm telling you to not trust Grandma. She's a devious old bird. Did you really believe that was her recipe for rhubarb pie? Riiiiiiight.
Many, if not most, records are going to contain a mix of primary and secondary information, so once again, you're going to have to really use that big, juicy chess club brain of yours to figure out which information fits into which category.
A death certificate, for example, is an original record, as it was created at or at about the time of the event (the death). The information on a death certificate, however, is a mix of primary and secondary information: information related to the death itself is primary, since those parts of the death certificate were usually filled out by the attending doctor who was an eyewitness. Information about the life of the deceased -- birthdate and place, occupation, family -- is almost always secondary, though. It, of necessity, came from someone other than the deceased, and so is secondhand information about his life and may be less reliable. This, however, does not hold true when the deceased is a child and the informant on the death certificate is one of his parents, since they were in a position of first-hand knowledge with regards to his birth and life. In fact, some of that information is even third-hand, as the informant may have been told by the deceased (when alive) that he was born in a given place, but, as we've discussed above, the deceased himself wasn't an eyewitness to his birth.
A death certificate informant may have been a family member... a spouse, or a child, or other relative, but also may have been a friend, a neighbor, a coworker, a staff member at the hospital, or the crossing guard who couldn't get to the drunk guy fast enough. In other words, you need to try to figure out who the informant was in order to judge how reliable the information on the record may be. Because the information giver on a birth certificate would have been a participant in the process, while an informant on a death certificate could have been almost anyone, a birth certificate or other birth record made at or at about the time of the birth will almost always be more reliable than a death record (a death certificate, obituary, gravestone, etc.) with regards to information about that same birth.
Or, for another example, let's return to the 1940 Census. In order to correctly analyze the information given, you need to know a little something about how the information was gathered. To really drill down into it, you'd want to read the Instructions to Enumerators for that census, available in PDF form here, but in case you're feeling lazy and just want the Cliff's Notes edition, here are a couple of salient points:
1) The enumerator would speak to whoever was present in the house; if the only person home was the 12 year old son, then the information in the census may have come from the 12 year old son, who wouldn't be a great source of primary information about his parents or grandparents.
2) If no one was home, sometimes the enumerator would speak to the neighbors, who would probably know even less than the 12 year old son.
3) Whoever provided it, the information on the census was written down by a third party who may have misheard your ancestor due to a thick accent, or may have gotten distracted and written a "2" instead of an "8," in someone's age, or may have written down your ancestor's answer to a question he or she misunderstood in the first place.
Clearly, there's rarely a way to know for sure exactly what happened, and exactly who provided what information to whom, but understanding the information-gathering process ought to help you see how much of a grain of salt you should be ingesting when evaluating whether the information you're seeing is really primary or secondary masquerading as primary.
Since this is all very technical, and, if you're like me, you don't have a very technical mind, let's switch things up a bit to help you keep these straight in your mind: let's refer to primary information as superhero information, and secondary information as vampire information. When you read a superhero comic, you're watching the brave and no doubt peerless superhero punch! and kapow! and splat! in real time (why, no, I've never seen the Adam West Batman... why do you ask?). You're not reading the newspaper article written later about the kerfuffle or watching the townspeople discuss what your caped crusader is doing; what you're observing is first-hand. When asked about it, you could offer an eyewitness account of superhero information. If you've ever read Dracula, on the other hand, you'll notice that, not only is none of the book from Dracula's point of view, but that it's an epistolary novel (that is to say, a novel written in the form of letters, memoranda, newspaper clippings and other written documents); as such, everything you're reading is what someone else observed about Dracula and his gluten-free preferences. If someone were to ask you, you could not give a first-hand account of the events taking place. You could only proffer vampire information.
Pictured: documentation of dubious origin.
So let's imagine a purely hypothetical man who was born in the late 1800s. Like many people of the age, he didn't really know how old he was (it may be shocking to us, since we all have multiple identifying documents with our birthdays on them and celebrate our birthdays each year, but birthdays weren't such a big deal back then, and your great-great-grandfather who was born in 1853 didn't have a driver's license in his wallet or a Social Security number in a folder at home). Throughout his life, his given birth year on various documents and records fluctuated between 1886, 1888, and 1890. How would one even begin to decide which records would be more reliable in attempting to nail down a likely birth year for him?
1) Determine whether the record is original or derivative. Are you seeing the actual record (or a faithful digital image of the record), or are you seeing someone's possibly unfaithful transcript of the record?
2) Determine when each record was created. In the event that the person tends towards a certain birth year in earlier records (IE, closer in time to the event) and then towards a later birth year in later records, the earlier records are more likely to be reliable.
3) Determine whether the information regarding his birth on the record is primary or secondary. Is there any chance that the record was being filled out by his parents? Or were all of the records based on his second-hand information? Or, as in the case of his gravestone, third-hand information from his wife or children who could only go off of what he had told them?
4) Determine whether he would have had a good reason to be less than truthful on any of these records. For example, back "in the day," some young men lied about their age in order to appear eligible for military service.
5) Incorporate other relevant information. When were his siblings born? Were any of them documented as being born at times that would invalidate one or two of the birth years he gave for himself?
Once you've done the final analysis and come up with the most likely answer (in this case, based upon the timing of the records and the birthdates of this hypothetical man's hypothetical siblings, the best educated guess is that he was born in 1886), go a step further and answer one more question: What record, if it exists, would you need to see in order to answer this question definitively? In the case of the man with the changing birthday, the only way to know for sure in which year he was born would be a birth record, either religious or civil, created at or at about the time of his birth.
Now imagine how hard this is if you're researching the birth of an actual vampire born in 1433 and who plagiarized many records just to, "fit in with the locals".
In the end, most of us are still left with questions about our families that we aren't able to definitively answer; often the original record we want no longer exists, or never existed in the first place. Doing critical analysis of the records we do have and the information in those records, however, will get us as close to the answers we seek as it is possible to get.
1 A genealogical index is a database that does not contain the actual record, but rather, indicates that the record you are looking for exists somewhere; it provides a citation to help you locate the records, and possibly additional transcribed information from the record. For example, see the below image of a page from the 1910 United States Federal Census from Ancestry.com: it contains both the digital image of the record itself, and the index of the record underneath.
2 An abstract is a written summary of the most important genealogical tidbits in a record, rather than a full and faithful transcription of the document.
3 I haven't actually bothered to do the research to find out what precisely would have actually been a scandal in the 1880s.