The Gap in the GRE

(For those who have better things to do than ponder GRE scores, this post will make more sense if you know that around four percent of all GRE testers achieve the highest score of 800 on the quant (math) section, while just 2-3% of all testers get over 700 on the verbal section.)

Razib Khan, building on his previous work, correlates GRE verbal and math scores by intended major into a stunningly cool graphic. Many commenters, both at Khan’s and Steve Sailer’s site observed the sizable gap between quant and verbal averages, repeated the amateur’s conventional wisdom that foreign testers, particularly Asians, are the cause.

This may be a small point, but could everyone please take note so they don’t irritate me with gormless speculation: verbal scores on standardized tests have been lower than math scores for forty years or more. High verbal scores are extremely rare; high math scores are, in comparison, common.

First up, Sex, Race, Ethnicity, and Performance on the GRE General Test provides GRE scores broken down by nationality for the year 2001-2002 (first year of the last change):

About 3% of US born testers get anywhere close to 700 on the verbal section. And just to forestall the next objection, no, it’s not URM scores that are dragging the average down, either:

(This is US citizens only).

In this year, at least, white women are 50% of the tested population; white men another 27%. This makes sense; most of the high volume, low academic barrier grad school specialities are white women jobs (teachers, social workers, nurses). That also explains the rather sizable gap between the genders; men who take the GRE are at least as likely to be testing into a hard sciences speciality as they are into teaching. But again, it’s clear that about 10-12% of whites or US testers are getting over 600 on the verbal.

This isn’t a recent development, as the score history from 1965 on shows:

(Source: NCES, and you can see all the scores through 2007 there).

From 1965-69, verbal and math scores had roughly the same average, although math had the greater standard deviation, which should mean that there were more 800s in math than in verbal. In the lax 70s, both verbal and math scores declined, although verbal scores dropped far more. In the “A Nation at Risk” 80s, math scores rebounded and exceeded the good old days (some of that growth, no doubt, attributable to the increased Asian presence). Verbal scores never did.

The GRE was originally a knockoff of the SAT, and the same decline in verbal scores can be seen through these years. Math scores didn’t take much of a dive during this time, interestingly enough.

None of this decline is news; Murray and Herrnstein’s Bell Curve first made the data available, I think. But it shows that long before Asians became overrepresented in college tests, verbal GRE scores have been low, and high verbal scores have always been rarer than math. You can see the same gap between verbal and math in the GMAT, and MCAT science scores are far higher than the verbal section. The LSAT doesn’t test math and I couldn’t find a breakdown of section scores. I coached the LSAT, though, and distinctly remember reading in the company manual (which I’ve since tossed) that, while most testers think the logic games was the most difficult, the reading section had the lowest average score. Can’t find any data confirming or denying that memory, unfortunately.

Why do we appear to have fewer high verbal achievers than math achievers? I think Murray and Herrnstein were correct when they wrote that “a politically compromised curriculum is less likely to sharpen the verbal skills of students than one that hews to standards of intellectual rigor and quality” annd that “when parents demanded higher standards, their schools introduced higher standards in the math curriculum that really were higher, and higher standards in the humanities and social sciences that really were not”. (Bell Curve, page 432-433) Without question, we have lost a couple generations of cognitively able students who weren’t given the opportunity to really achieve to their fullest capability, and we stand to lose a few more.

But I also wonder if verbal intelligence is less understood and consequently less valued. If one is “good at math”, there’s a logical progression of courses to take, problems to solve (or spend a lifetime trying to), and increasingly difficult subjects to tackle–and plenty of careers that want them. But if one has a high verbal intelligence without good spatial aptitude (which seems to be necessary for higher math) it is often described as “good at reading”, a woefully inaccurate characterization of high verbal intelligence—and then what? Apart from law, there aren’t nearly as many clearly defined career paths with a wide range of opportunities for all temperaments and interests. Most of the ones I can think of involve luck and driving ambition just to get started (journalism, tenured academia, political consultant).

For a good twenty years or so, people with high verbal skills who were indifferent at high-level math went into technology. It’s hard to remember now in the age of Google and after the heyday of corporate computing, but IBM and mainframe shops were filled with bright people who had degrees in history and English and humanities who just “didn’t like math” but were excellent programmers. I routinely worked in shops where all the expert techies making six figures came from non-STEM majors. But that time appears to be over.

Of course, doing anything about this lack of clearly defined career paths for smart folks with less spatial aptitude would involve acknowledging it’s a problem, and I might be the only one who thinks it’s a problem.


About educationrealist

33 responses to “The Gap in the GRE

  • Violet

    While verbal scores are only slightly different between U.S. Citizens and non-U.S. citizens, the latter have higher scores in non-language parts (considerable difference in Quantitative). From Razib’s posts, higher total scores imply higher I.Q.s. Then, why are these higher I.Q. non-citizen participants performing poorly on verbal?

    Shouldn’t we compare equivalent I.Q. native speaker performance to see if the verbal difficulty is general or specific to non-native speaker?

    Regardless of the teaching problems in U.S. causing lower verbal scores for all U.S. citizens, there is still an inherent advantage in performing a reading comprehension task in one’s native language compared to one’s second language, especially one that is learned as an adult. I would have thought that is obvious.

    Therefore, the total GRE score is under-estimating the non-native speaker’s I.Q. These stats aren’t convincing at all to see if high math-low verbal scores in some fields are due to relatively higher proportion of non-native speakers specific to those fields.

    • educationrealist

      “Shouldn’t we compare equivalent I.Q. native speaker performance to see if the verbal difficulty is general or specific to non-native speaker?”

      We have. Asians have lower verbal IQs in their native language than whites do. However, that doesn’t have anything to do with the primary point of this piece.

      This does, though: ” These stats aren’t convincing at all to see if high math-low verbal scores in some fields are due to relatively higher proportion of non-native speakers specific to those fields.”

      Whites are not routinely scoring over 700 on the Verbal portion of the GRE. Full stop. Asian Americans, who have lower verbal GRE average scores than whites, are also not routinely scoring over 700. International testers have a wider standard deviation, so the score at 2SD is about 690–much the same as US citizens and US whites.

      So non-native Asians aren’t dragging down the verbal scores. Neither are Asian Americans.

      An interesting question is whether or not non-native Asians would drag down scores if they weren’t cheating in such huge numbers, but that’s another story.

  • ohwilleke

    “Why do we appear to have fewer high verbal achievers than math achievers?”

    The knowledge base from which the high end of the verbal section draws is much larger than the knowledge base from which the high end of the math section draws.

    The math section deliberately censors out materials taught only to math students from the GRE exam taken by all undergraduates so as to not test non-technical majors on something that they have never been taught and reserves the high end material for a subject matter test. So, anyone who has a perfect mastery of high school math is in a position to ace or nearly ace the GRE math. For a math major like myself, that meant getting a 780 of the GRE math was no big deal even though I was only a middling B+/A- student in my major.

    The verbal shows no such mercy.

    • educationrealist

      But don’t you think that’s a symptom of exactly what I’m saying, that no one takes verbal intelligence seriously? The testers don’t think of verbal subjects being something to master with study or knowledge. That’s true, of course, but what they don’t take into account is part two–that the high end is way, way out there.

      And still, some people achieve that high end in verbal, and the achievement has little correlation with years of study or subjects studied. Meanwhile, there’s no analogous situation in math. The test can be made difficult using just algebra and geometry, but a high score is achievable by far more people–even though even that small amount of math requires more years of formal study than the verbal does.

      So we’re still back at the fact that fewer people are found out in the stratosphere in verbal abilities than mathematical.

      • EH

        I’m one of those with the ceiling verbal scores – 780 on the mid-’80s SAT, 98 MAT (practice), 180 Schmies vocabulary (a fun little online test, 200 synonym / antonym pairs only takes about 20 minutes).

        Since they took analogies out of the SAT, and added more of those awful “which best expresses the author’s intent?” sort of questions, the test has less accuracy and “top”. Personally, I blame the dyslexic ex-West Virginia governor /insurance company scion who was running ETS at the time. At the same time the math has gotten easier, but I’ve kept up my use of math over the years, so perhaps it’s just me. The GRE math subject test is quite a different matter, starting off with group theory and moving out into stuff that would send even Lovecraft’s many-angled ones fleeing in terror. Some of its questions have below-chance correct response rates even for math majors. The computer science subject test is nearly as bad. On the other hand, the psychology GRE can be passed by doing little more than skimming the tables of contents of a couple of textbooks before the exam.

        So far as I can tell verbal ability is mostly good for reading. Writing, Scrabble, conversation – not so much. But then again, reading is more reliable and varied pleasure than anything else in life.


        I had seen that second table you used in your essay several years ago when looking through ETS research papers but was unable to find it later.

        Note the Asian men / Black women gap: 653 vs. 407 !!
        The average Asian man is at the 98th percentile for Black women. The average Black woman is at the 2nd percentile for Asian men.

        Also note the larger standard deviations for men than women except in verbal scores and in the Asian quantitative scores. The exact size of the difference in standard deviations between men and women and by race is difficult to come by.

  • Steve Sailer

    I wonder if American culture — relative to British culture — is deficient at educating us verbally. I’ve been convinced since high school that English writers were better than American writers.

    I was recently re-reading “Great Contemporaries,” a collection of popular journalism Winston Churchill wrote (or, to be precise, dictated) in the 1930s about celebrities he’d known. For mastery of English, for vast and precise vocabulary, I can’t imagine any American politician of the last century coming close. Teddy Roosevelt had comparable mental energy, but nobody reads his books for fun these days.

    • educationrealist

      I agree with you about writing, but with the exception of the Jesuits, Americans just don’t find composition all that critical. So we don’t teach writing very well. I didn’t really learn to think about the quality of my writing until my mid-30s. Part of this, I think, is because English teachers in America have, for 40 years or more, focused more on literature analysis than composition.

      But I don’t think that strong verbal skills have much to do with writing, except incidentally. I’ve had really strong verbal skills since birth, with a college level vocabulary by age 8 and writing has always been incidental to that strength. It’s more about information retention, content acquisition and–if combined with strong logical aptitude–analytical capabilities.

      Where are the scholarships and high school prizes for great history analysis, or written persuasion? If we created incentives for students to show their analytical skills and improve their writing, the best would rise to the top. I think the whole area needs the equivalent of the Westinghouse competition.

      That would also create more logical career paths–that, to me, is a critical missing area.

      Are there any tests like the GRE or SAT in Britain that we could use to check the score distributions between quant and verbal?

  • random mutation

    “But again, it’s clear that about 10-12% of whites or US testers”

    Was that takers of tests or givers of tests?

    I have noticed that women pay attention to verbal skills among men.

    Further, the numbers you present above are interesting because while there are fewer men taking the GRE than women in all categories (and I agree with your explanation) men taking the GRE have higher average verbal scores than women taking the GRE. Of course males have higher average scores in all categories, but what is interesting here is that it seems to contradict the perception that women are more capable, verbally. (Although, since the numbers suggest that the GRE is reaching further down the scale among females, it is perhaps not so surprising.)

    Why is this? Is it that those males who are less intelligent in any area are more realistic about the barriers for men and simply don’t bother about higher degrees?

    (It is curious, as well, that among Asian/Pacific Islanders, who seem to be mostly East Asian judging by their math scores, the female SD is larger than the male SD.)

    • educationrealist

      but what is interesting here is that it seems to contradict the perception that women are more capable, verbally.

      The “women are more capable verbally” comes from two key areas, I think. First, girls (not sure about women) are much more expressive and detailed in their writing than boys are, and they routinely outscore boys on writing tests. The NAEP reading test always shows a big gap, but the NAEP is actually more a writing test than a reading test (see here for examples), which casts all of NAEP’s findings into doubt (I am not a fan).

      But if you look at SAT/ACT scores by gender, particularly in states like Illinois and Colorado, who give all high school seniors the ACT, you find that the reading scores are close to equal–it’s usually the smallest gap of the four scores. On the SAT boys outscore girls, but that’s due in part to the low-ability boys having been left behind–they have more behavior problems; low ability girls are more likely to finish school and thus take the SAT.

      Is it that those males who are less intelligent in any area are more realistic about the barriers for men and simply don’t bother about higher degrees?

      I think it’s more about career options and the way that education developed. You want to become a plumber or a general contractor, you don’t go to college. You want to become a kindergarten teacher, a child-care worker, an occupational therapist or a social worker, you go to college. This difference has nothing to do with the intellectual skills of one job or the other, but rather the ways in which the career training developed.

      For this reason, GRE mid-range scores are pretty close to useless for determining any sort of ability distribution. Many master’s degrees are intellectually identical to many trade skills. It’s not that men see barriers, it’s that men are more likely to want to be plumbers.

  • random mutation

    “In the lax 70s, both verbal and math scores declined, although verbal scores dropped far more. In the “A Nation at Risk” 80s, math scores rebounded and exceeded the good old days (some of that growth, no doubt, attributable to the increased Asian presence). Verbal scores never did. ”

    While I do not disagree with what you say about verbal vs math skills, isn’t the general decline a function of the fact that an increasing percentage of people with bachelors degrees were taking the test, and thus there were more less-capable people taking the test and pulling the average down? Even the rebound in the ’80s seems to track the decline in the percentage taking the test during that period?

  • KLO

    Language is something that we interact with every day to varying degrees. It is not first learned in school, but rather at home. This learning process begins at birth and continues until the day we die. People vary greatly in the amount of exposure they get to language. Some people are voracious readers, devouring literature from an early age. Other people enjoy largely non-linguistic activities, and have significantly less exposure to language.

    Moreover, one’s ability with language plays no small part in determining one’s the level of exposure. People who are good writers or readers will get much more exposure than people who are not good writers or readers. Thus, relatively small gaps in innate ability will tend to flourish into wide chasms over time.

    Math is quite different. Very few people solve complex math problems for fun. For most everyone, math is something learned in school. The amount of time spent with math will not vary much from person to person until one decides on a career path. Sure, some people will spend more time on math than others, but the differences will nonetheless be quite small.

    Why, then, is there greater disparity among people in their ability to use and understand language than there is in their ability to use and understand math? It is simple. Within a given population, the amount of time and energy devoted to language will vary enormously. A not inconsiderable number of people will read for hours each day. Another large group of people may not even read for 15 minutes each day. Over time, the differences in exposure will lead to enormous differences in ability. The same is not true of math, where the differences in exposure are orders of magnitude lower.

    It is difficult to overcome the differences in linguistic ability through dedicated study for two basic reasons. First, the number of hours of practice needed to equal the number of hours of exposure an elite language user has built up over time is enormous. There are simply not enough hours in the day to catch up over a time period shorter than several years. Second, elite language users are continuously exposed to high level language for many hours each day. Once a person is behind, there can be no catching up. The only hope is to maintain one’s degree of inferiority.

    Think of just how few elite writers of the English language there are who did not learn English at a very early age. Even Nabokov, who was an excellent writer in both English and Russian, learned English as a child. The only person of whom I am aware who managed to write great works in English despite not having learned English as a child is Joseph Brodsky. Can you name any others?

    I don’t think that early exposure to math is quite so critical. A person who does not learn much math by age 13 can pick up it quickly enough to do amazing things with it. Language is a much harder proposition.

    • surfer

      Joseph Conrad. Also, I’m not sure that being a famous writer is the most useful measuring stick. But I like your ideas on language learning.

  • Learning Math « educationrealist

    […] to write an analytical essay on Hamlet and offer some original insights. Unfortunately, as has been noted before, our history and English curriculum has never recovered from the dumbing down it suffered through […]

  • Cornelius

    But if one has a high verbal intelligence without good spatial aptitude (which seems to be necessary for higher math)…

    What’s your definition of good spatial aptitude?

    Jews have a unique intelligence profile with spatial intelligence significantly lagging verbal intelligence, but they still make significant contributions to higher maths. I don’t think we can assume that all these Jewish mathematicians are those who do not share the unique intelligence profile of their kin.

    Look at Jewish performance on the Putnam exam.

    Probably most areas of number theory rely more on what we would call verbal-analytic intelligence than on spatial intelligence.

    • educationrealist

      I agree. And I’m much better at number theory than spatial, for what it’s worth.

      But remember, there are an infinite number of combinations. It’s not like High V, Low S, Mid L.

      I wrote more about this here–it’s a lot longer, though, and personal. But at the end it goes into more of this as well.

      ONe thing it’s always essential to remember about my cog ability posts–I’m not an expert, and I’m talking at something like 90,000 feet. Sketching out ideas, not fixing it frame by frame.

      Learning math:

  • 2012 in review « educationrealist

    […] The Gap in the GRE, currently 8th in my overall list, has actually gotten more views over time than in its original posting. Steve Sailer discussed it in June, and it’s getting more interest over time. As a 99.999% verbal performer, I’m proud to have increased awareness of that gap. […]

  • Steve Benton

    “I’m not suggesting that all Asians cheat, nor do I think that high Asian American scores are due to cheating…” Wow, that is such a racist comment. It’s like saying “I’m not suggesting all middle-class, suburban white kids are spoiled, self-entitled brats.” Although many are most blacks, Asians, Latinos don’t walk around making this assumption about every suburban white teenager that we meet.

    • educationrealist

      a) since I’m NOT suggesting it, what’s your problem?
      b) cheating is a known issue in schools with large Asian populations. And that means all forms of Asians except Pacific islanders.

      There is a cultural dimension to cheating, or rather a motivational aspect to cheating that has a cultural dimension. Low ability kids who cheat do it as a survival strategy, not an attempt to pass themselves off as better than they are. High ability kids who cheat–and again, that is a disproportionately Asian crowd–do indeed want to make themselves look better than they think their uncheated results will show.

      I’m uninterested in whether I’ve offended your sensibilities.

  • namae nanka

    “From 1965-69, verbal and math scores had roughly the same average, although math had the greater standard deviation, which should mean that there were more 800s in math than in verbal.”

    not necessarily.

    “long before Asians became overrepresented in college tests, verbal GRE scores have been low”

    SAT-M scores used to be lower once upon a time.

    “But anyone who thinks that memorization is a key factor in high verbal scores isn’t acknowledging that we’ve got a demographic willing to put in the time, and it’s not made a big difference. ”

    how do you know that asians aren’t overrepresented at higher verbal scores? The average poses problem with the mixture of different ethnicities.

    Perhaps preparation was a wrong term to use, especially its confounding with what has become the asian test prep industry. What I meant was fluid vs. crystallized.

    I haven’t taken any of the three, had friends who took the GRE and concentrated on mugging up obscure words since the maths portion was pedestrian. The only perfect scorer I heard of was the brother of a very smart friend.

    “Why do we appear to have fewer high verbal achievers than math achievers?”

    ‘Finally, an interesting trend was revealed. The presence of exceptionally high verbal ability appeared to increase the likelihood of the presence of high mathematical ability. Only one of the verbally precocious students had an SAT-M score lower than 500 (the average score of a college-bound 12th-grade male). The reverse was not apparent: high mathematical ability did not seem to indicate concomitantly high verbal ability. ‘

    From SMPY, not sure how much of this was an artifact of asians, asians are the majority now, asian females have been majority since the 80s.

    Agree with lower verbal preparation today than 50 years ago, perhaps philosophy instead of literature would be a better way to bring out the nuances of language.

    and any idea what lead to this?

  • Luke Lea

    Nice post. Where do you think these high verbals are ending up in our society? Are we absorbing them?

    • educationrealist

      As I said in the essay, I think a lot of them originally went into computers, back in the 70s and 80s. I’m not sure where they’re going now. Certainly not into tech, and corporate IT is a shadow of its old self. In recent years (say the last 15), smart kids who weren’t instantly good at math have gotten the message that they better *get* good at it, so some of them probably meander into business. But it’s an interesting question, and I don’t know where the younger generation, really good at verbal but less strong in math, are now.

      (Of course, I stayed in tech and then went into teaching—and blogging. Maybe I’m not all that unique!)

  • Anonyias

    In response to both Luke and the post: I’m a relatively “high verbal.” I scored in the 94th percentile on the verbal GRE (without studying), and perfectly on the Reading/English sections of the ACT. I was decent at Math as a child but gradually lost interest. I’ve never had real trouble learning required Maths, yet it’s still clear that I lack natural aptitude in that area.

    I’m attempting to become a secondary Social Studies teacher, but I’ve heard the job market is horrible. I opted for this path after receiving numerous warnings about my first career choice, law.Though I love student teaching, I still feel somewhat aimless.

    I don’t think American culture values verbal intelligence. Its usefulness seems to subside with age. Maybe it is my geographic location, but a large vocabulary both irritates and alienates people. And then there is the perception issue: a lot of people charge that all there is to verbal intelligence is a good memory. All of it depresses me. I would much rather be mathematically gifted and verbally so-so.

    • educationrealist

      I don’t know if the job market is horrible. It’s not easy, though, certainly not for history and English. Learn how to interview for teacher jobs–I have a few posts on that, but they mostly consist of What They Want, not What to Say.

      It’s funny, my students constantly point out that I use “big words”. I’m always like, what?

      Sorry you feel out of kilter. Good luck.

  • Jon Kessler

    Are there any numbers that specifically break out the results for US-born test-takers of each ethnicity? I wonder if being exposed to the American cultural milieu from the git-go has any kind of effect, whether positive or negative.

  • Robert Evans

    “But if one has a high verbal intelligence without good spatial aptitude (which seems to be necessary for higher math) it is often described as “good at reading” ”

    Yo man,

    The highest purely “verbal” scores I’ve ever had were 670 pre-95 SAT, 660 2010 GRE, and a 29 ACT (twice). Compare that to the 33 and 35 Reading ACT subscores, and an ability to devour a 300 page fiction book between walking to the bus stop in the morning and getting home from school in the afternoon when I was a child.

    I’m good at math and good at reading. My wife is the person I turn to when I need help writing or with vocabulary.

  • College Admissions, Race, and Unintended Consequences | educationrealist

    […] Have whites and Asians have gotten a lot smarter in verbal? Um. Yeah. Not. If you won’t take my word for it, check out GRE scores during that time, which was very similar to the 1995 SAT throughout the 90s and before and after, did not see a corresponding increase in scores. […]

  • surfer

    I seem to dimly remember that c. early 90s, the GRE math section was actually easier questions than the SAT. Or the Princeton Review prep book that I had said something like that.

  • Benjamin

    Would you trust the gender standard deviations and averages to give an accurate estimate of gender differences at the far right tail. Some calculations suggest that there are a ton more men than women at the far right of the verbal curve. I’m not so sure that’s right.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: