Line #4 UsefulNotes/IQTesting


1	Intelligence Quotient scores and Intelligence Quotient testing do not work the way that they are depicted in most fiction. Since there are [[ImprobablyHighIQ a couple]] [[ImprobablyLowIQ of tropes]], both dealing with the two most common (and unfortunately flawed) portrayals of these, here's the lowdown on how IQ actually works.
2
3	!!!Accuracy
4	There is still much debate about what an [=IQ=] test actually measures, whether it is actually a good measure of intelligence, a measure of only a part or one type of intelligence, or actually a measure of test taking skills. TV assumes that if [=IQ=] tests don't measure intelligence 100% accurately, they're at least measuring something basic and unchanging about a person. Research suggests not: intelligence seems to change over a person's lifetime, and certain types of training help people boost their [=IQ.=] However, it should be noted that this is a shortcoming of the [=IQ=] tests used to measure intelligence. Lastly, [=IQ=] scores have seen something called the ''Flynn effect'': a steady rise over the past half-century, slowly but across the board (although possibly now dying out). To keep 100 as the mean, the tests themselves have to be made harder.
5
6	Additionally, social, psychological, and cultural factors can often skew a score. For example, some studies have suggested that African-Americans score better on [=IQ=] tests when they think their score is only going to be compared to those of other African-Americans. In regards to a culture skew, many [=IQ=] tests include a core measure of a subject's general knowledge of things such as famous historical people/events (e.g., Who was Pocahontas?) or social norms/mores (e.g., What are you supposed to do if you find a wallet on the street?). Many professionals critique this as being culturally biased in favor of White/Caucasian culture to the detriment of not only minorities, but also immigrants.
7
8	[=IQ=] tests are intended to measure a number of areas of intelligence, not just "can do mental arithmetic very fast". Mental arithmetic is a skill, that (like, say, juggling) can be taught, and has more to do with memorizing tricks and log tables than innate intelligence. Pattern recognition is emphasized over math skills in most tests. There's also a hotly-debated theory that there are ''[[http://en.wikipedia.org/wiki/Theory_of_multiple_intelligences many]]'' kinds of specialized intelligences, some of which an IQ test may not even bother assessing. IQ tests may also require skills that people with brain injuries may have lost but which don't actually affect basic intelligence, such as the ability to do math in one's head or to understand perspective. Finally, a full-scale IQ number does not tell people which areas a person excelled in: A person with an average-range IQ may have performed exceptionally in the verbal reasoning and comprehension portion, but bombed the working memory or processing speed portions.
9
10	This takes us to the next point: with any tool--but ''especially'' with psychological tests--you need to be aware of what the thing is trying to measure, and whether it actually accomplishes what it claims to. The original IQ test, the Stanford-Binet, was designed simply to identify what grade of school you ''should'' be in, versus which grade you actually were, with an eye towards identifying kids with developmental disabilities. In other words, your Stanford-Binet IQ number gets smaller every year, as you progress through numerically-larger grades, and stops being meaningful the moment you get your diploma. Obviously, a test that does ''not'' calculate using school grades as a yardstick (eg. the [[http://en.wikipedia.org/wiki/Wechsler_Adult_Intelligence_Scale Wechsler Adult Intelligence Scale]] or more recent versions of the Stanford-Binet) will not have this problem... But how does ''it'' measure? It turns out you can't give the [=WAIS=] to a child and expect meaningful results, and there's a "Wechsler Scale for Children" for that exact reason. Psychological tests are designed to be used under ''very'' specific circumstances, and if you move outside them, the test and its results may not be applicable at all anymore.
11
12	!!!Range
13	ImprobablyHighIQ and ImprobablyLowIQ are staples of the writer's trade; in reality, [=IQs=] range somewhere between 50 and 200. While in theory the average (statistical mean) IQ is 100, in practice the average of people in the street tends to be slightly higher because people with IQ scores under 70 are typically under various levels of care or supervision, though not by much, since the vast majority of people fall into the "average" range. Additionally, due to cognitive sorting, people tend to hang around with other people of similar intelligence so it would not be uncommon for a college grad with an IQ of 120 to have friends who average about the same.
14
15	[=IQ=]s were originally given as ratios: (mental age / chronological age) * 100 = IQ. Thus a 6 year old who scored as well as the average 9 year old would have an IQ of 150 (9 / 6 * 100). This system matches contemporary IQ scores quite well--until you get more than about three standard deviations from the mean, after which point ratio IQ scores stop being distributed on a normal curve. Thus, if IQ is meant in the original sense, then it's at least mathematically possible that a character could have an IQ of 300: they would simply have had to have been functioning at the level of a twelve-year-old when they were four. At any rate, [=IQ=]s haven't been ratio-based for sixty years.
16
17	[=IQ=] test scores are designed to follow a normal (bell curve) distribution--meaning that, if an IQ test is normed at 100 and has a standard deviation of 15 points, about 68% of the population has an IQ between 85 and 115 (one standard deviation from the norm), and fully ''95%'' of people are between 70 and 130. UsefulNotes/{{Mensa}}, the best-known international society for people with high [=IQs=], requires a score of at least 132 on the Stanford-Binet or Wechsler tests, corresponding to the 98th percentile. [=IQs=] over 145/under 55 number about one in a thousand; [=IQs=] over 160/under 40 (four standard deviations from the norm), about one in ''thirty'' thousand. The occurence of the more ridiculously high levels that come up in fiction become difficult to calculate, even in theory, but [=IQs=] over 190 (6 standard deviations from the norm) are about one in a billion, while [=IQs=] over 229 (8.6 standard deviations from the norm) would be about one in ten trillion: an interstellar level of IQ. For more numerical stats, see the analysis pages of Analysis/ImprobablyHighIQ or Analysis/ImprobablyLowIQ.
18
19	That far from the middle, test makers have trouble finding enough people to produce a nice reliable sample. Even then, and even assuming such a person had nothing better to do than help psychiatrists norm their IQ tests, the same person would likely achieve slightly different scores each time due to differences in the tests, the specific questions, and the conditions under which the test was taken.
20
21	Measured IQ, like most types of human variation, does not exactly follow a normal distribution in practice. The most obvious deviation is that very low scores are much more common than very high scores, because measured IQ can be greatly lowered but not raised by various disorders and traumas affecting the brain.
22
23	It's also not actually certain whether the distribution of human intelligence actually ''is'' in the form of a bell curve. If it's not, then formulating IQ tests specifically to create such a curve means that inaccuracy is being built into them.
24
25	!!!Precision
26	A test that can successfully classify 95% of the population will be adequate for nearly all conceivable uses... but eventually it will run into a ceiling. If two people get a perfect score, the test provides no way to measure [[RankedByIQ which of them is more intelligent]]. A more difficult test must be used, and this is not commonly done. Mensa and other high [=IQ=] societies are some of the few groups that are interested in quantifying IQ scores at the extreme high end of the range. Since a special test must be used to give an accurate and precise score for these individuals, and since most people never need to do this, it's common for people with high intelligence to not know their exact [=IQ=] score beyond a fuzzy range.
27
28	Similarly, if two people both get every single question wrong (or get the number of correct answers you'd typically get from randomly guessing), the test does not tell you which one is less intelligent. Most standard IQ tests only go down to 40 or 50 IQ at the lowest. Unlike high IQ, however, there are plenty of people interested in specialized tests able to measure the lower ranges, since this information is important for special education programs and other kinds of services. As a result, several tests are designed to be highly specific for [=IQ=]s under 50. There are also adaptive behavior tests used to determine what practical self-care skills the person has, which can estimate IQ in someone who is difficult to test.
29
30	!!!The Score
31	People in TV shows never ever qualify their [=IQ=] tests in any way. This is akin to saying you got a 36 on the college admissions test without saying which one it is; a 36 could mean perfect score, top 75%, or below the lowest possible score, depending on whether you're talking about the [=ACT=], the International Baccalaureate, or the [=SAT=] respectively. Just giving a score assumes that all [=IQ=] tests have the same standard deviation, or measure the same kinds of brainpower. They don't. Tests have standard deviations ranging from 10 to 20 or higher; for example a score of 132 on the Stanford-Binet is equivalent to a 148 on the Cattell.
32
33	This may be partially justified by the use of the term [=IQ=] itself, which ''[[DependingOnTheWriter can]]'' automatically imply the bell curve/mean = 100/standard deviation = 15 pattern described above. What is certainly true is that raw test scores from different tests cannot be compared in any meaningful way; only two transformed values based on said specific bell curve pattern could even come close to making sense in comparison to one another. Also, different tests have been normed to different populations, meaning that differences in their norming samples might make even such values basically incompatible.
34
35	In addition, most IQ tests have at least two components, if nothing else, something separating an equivalent to verbal IQ (language and decoding) from performance IQ (perceptual abilities and/or executive function). It is not unheard of for someone taking an IQ test to come out at opposite extremes of the two scales, leaving the composite (total) score worthless for informational purposes - in such a case the composite score shows an average IQ, when the useful information is a very high score on one part and a very low one on the other.
36
37	!!!Just To Make Things Worse...
38	Although the term "[=IQ=]" implies that a test has been taken, it is possible for psychologists and psychiatrists to ''estimate'' a person's [=IQ=] without ever administering a formal [=IQ=] ''test.'' The famous Rorschach test ("inkblot test") is less concerned with measuring what the person thinks the shape looks like but ''how'' they think it looks like, which provides insight into a person's thought processes. The [=TAT=] asks a patient to put a series of simple sketches in chronological order and then create a narrative. These tests, along with many others, often provide sufficient insight into a person's thought processes and level of reasoning to allow an expert to estimate the subject's [=IQ=] number.
39
40	Of course, the Rorschach test is one of the most contested tests, with multiple analysis sets and training. Two researchers who have been trained to analyze responses in the same way will likely draw the same conclusions, but their results may not match those of another analytic model. That being said, that makes the Rorschach about as a reliable as most conventional [=IQ=] tests, and since it's often administered as part of a package, it allows a sufficiently trained researcher to make a credible guess as to someone's [=IQ=]. This method is rarely shown on [=TV=], as [[ThereAreNoPsychologists no one needs therapy]] and it's hardly as impressive as having a character simply rattle off a concrete number.[[note]]Also:The public display of actual Rorschach images -- either on TV or described in text -- is considered an enormous taboo, as much of the Rorschach's utility is removed if the subject has already been exposed to the images.[[/note]]
41
42	However, all processes still face the same hurdles: controlling for [[http://en.wikipedia.org/wiki/Stereotype_threat stereotype threat]]; validity (does the test measure what it says it measures); reliability (does the test consistently yield the same results); and a sufficiently large sample size to produce meaningful results. Whether or not human intelligence can be accurately plotted as an [=IQ=] number is unclear. However intelligence is defined, even tests that measure that definition of intelligence are going to have to overcome all the hurdles previously discussed. There is also a sizable language gap in how [=TV=] writers and psychologists (et al) define [=IQ=], nevermind the fact that researchers can compare how a person does on multiple tests and derive multiple [=IQ=] numbers.
43
44	In short, the result that matters is determined by what the test is being used for: a student who significantly deviates from the mean on an [=IQ=] test is being underserved by their school in some respect, with focus less on the number than on the degrees of separation from the mean; a criminal whose [=IQ=] is significantly low, to the point that there is doubt as to their ability to form criminal intent, will need to be tested extensively, as inability to perform higher-order logical reasoning processes is critical when determining eligibility for the death penalty in the [=USA=]; and then there's Mensa and other private companies that will offer to tell you your number for a fee, if you truly want to know a "number."

Context UsefulNotes / IQTesting