One of my friends posted one of those unofficial internet IQ tests on Facebook a while ago. He got a ludicrously high number -- totally not surprised at that, he's practically the poster boy for ADHD genius. He gets particularly Sherlocky when drunk, as it makes him slow the fuck down. I have a long, proud history myself of taking those things at parties, when I'm completely obliterated. I consistently test out at 140 when I've had so many daiquiris I'm leaning my head on my hands and reading the screen with one eye closed to make the cursor stop wobbling.

I know a lot of people who do this, and the reason we take it about as seriously as your average OK Cupid quiz is that scores in this range are meaningless. Anything over two standard deviations -- 130, give or take about five points, depending on which IQ test you're using -- essentially just equates to "thar be smarts". There's no particularly good agreement on where the cutoff for "genius" is, but most of the GATE (gifted and talented education) programs I've heard of will automatically accept students who test out at 120 or higher.

(If you just said, "...the line is that low?" congratulations, you're probably in my bracket. We have LEGO over there on the table, a tub of oobleck in the corner, Top Gear on the TV, and free wifi everywhere in the building. Whiteboards and teaching spaces are available for when you inevitably need to share your cool brain-toys with other people.)

The main problem with these tests is that you can only control the accuracy of your results to the extent that you can control the accuracy of your test environment. When your test environment is the human brain, all bets are off.

The point of an IQ test is to assess your ability to take in information, analyse a situation, and come to logical conclusions about it. (We will leave aside the question of whether this is the only, or even the primary, factor in what other people perceive as intelligence. I'm not writing a thesis here.) In order for them to do this, first they have to be sure you understand the question. If you only speak English and they hand you a test packet entirely in Swahili, in the words of the immortal Bender: "You're boned." They try to make these things value- and culture-neutral, insofar as you can do that, but there are constant arguments over wording and content that mostly boil down to, "The IQ scores for [group X] are artificially low because your test questions are making [cultural assumption Y] that they do not share." For more on this problem, see any one of about four hojillion very long and often very ranty books on the topic. Amazon is your friend.

Okay, so what happens at the other end? Why are the high scores so scattershot?

Part of your score for these things is, obviously, whether you get the right answer. (We are also going to ignore the problems that a rigid definition of "right answer" can cause. Every gifted kid has stories of getting dinged on a project grade because they did something creative and whoever was scoring it failed to follow. Personally, it drives me bats when I get a foreign language teacher who can't tell the difference between me screwing up at grammar, and me screwing around with grammar.) The other part of your score is how long it took. Your time for a given question is from the instant they give it to you to the instant you give them your solution -- including the time it takes you to read the thing and stuff all the pertinent information into your noggin.

The people who do well on IQ tests tend to also be the people who used to fight their siblings for the cereal box at breakfast, because having to sit at the table with nothing to read would make their brains implode. The more you read, the faster you get, mainly because you develop what's called a "recognition vocabulary". Your recognition vocabulary is the collection of words you recognize as a shape unto themselves, without having to actually scan them letter by letter. Instead of going, "D... O... G... let me check the filing cabinet here... aha, this must be DOG!" your brain just takes in the word as a whole and goes, "DOG! I've got that right here!" When you spend every second of every day scanning your environment for something to read, if not just jamming your nose overtly into a book while walking around, your recognition vocabulary can get quite large and varied. Not having to go through word analysis for most of the English (or whatever) language knocks huge chunks off your time on an IQ test, raising your score by a lot.

(On the downside, you spend a lot of irritating seconds of your life staring blankly at pieces of paper, because you've learned the hard way that if you hand it back when you're really done with it, the other person will accuse you of not having actually read it.)

This is why none of the IQ and IQ-type tests they gave me as a kid gave them any kind of sensible result. I happen to be wired for language. I started reading signs aloud to people when I was about a year and a half old; God only knows how long I'd been doing that and not bothering to share. I was maybe four before they figured out I could read grown-up cursive upside-down. My home life was crazy and I had exactly zero real friends until I was a teenager, so I coped by living inside books. All of their score sheets just went TILT, because there is no such thing as a test that is calibrated properly for a kindergartener who is reading at a 12th grade level. Likewise, there's no such thing as a test that is calibrated properly for an adult who isn't reading sequentially so much as just jamming snapshots of the page directly into their brain. It becomes a case of taking the limit as reading time approaches zero. A little jog left or right on an asymptotic function like that can change your Y value by a lot, so to a very real extent, when you are in the high upper reaches, your exact score depends on how many of the words on your version of the test you are particularly close friends with.

This idea generalizes quite well to the rest of the test. IQ tests are, by design, made of basic spatial and reasoning puzzles. They are stripped down as much as possible in an effort to eliminate confounding variables like presentation or cultural connotation or just plain distraction. They just want to know if you can spot patterns. Gifted kids are good at spotting patterns -- so good, in fact, that it's not uncommon to look at one of those things and go, "Hey, I've already seen a brainteaser shaped just like that!" Essentially, you develop a recognition vocabulary of puzzle bits just like you do with words. When your time-per-question is already extremely low, recognizing whether or not you've already seen that one in a back issue of Scientific American can knock you a couple seconds, and hence points, in either direction.

I'm not saying that the ability to pull this shit out of hammerspace isn't an aspect of genius -- it most certainly is -- but in order to use the skill you must previously have run into that particular bit of relevant information somewhere outside of the test environment, and they have no way to control for that. They're trying to test for reasoning ability independent of prior knowledge, but they can't keep you from cheating when the cheat sheet is in your brain, so it's an inevitable bit of confounding slop. It just happens that because of the aysmptotic nature of the bell curve function, once you get out onto the thinned-out ends, little bits of slop add up quickly.

Depending on the test, your age, the environment, the time of day, the phase of the moon, how drunk you are, whether you get a comfortable chair, how many of these damn things you've done already, etc. etc. etc., one person can easily get a range of scores that bounce around by as much as twenty points. As a practical matter, once you're out past the 98th percentile, your exact score means bupkis -- you're out in the area where people don't give you funny looks because you know stuff, they give you funny looks because you think weird. You're starting to develop your own internal organizational schemata that, to the casual outside observer, differ from schizophrenia mainly by a lack of tinfoil haberdashery.

(Also as a practical matter, all of the people in the discussion that prompted this are 98th percentilers. People who break the tests generally don't think much of them. It's been my experience that the people most concerned with my test scores were people who were squarely average and wanted an excuse for not knowing what the hell they were supposed to do with me. Gifted kids tend to pass IQ numbers around as silly trivia, like Meyers-Briggs personality types. It's kind of like horoscopes for science nerds.)