Implicit Learning

The intelligence test is badly named. The main problem is that we should be talking about intelligence tests in plural, so that the IQ test is merely one of the many measures we use to assess our innate mental skills.

Unfortunately, despite the best efforts of Howard Gardner, Robert Sternberg and others, the IQ test remains the singular test of individual cognitive ability. The mysterious entity that it measures - g, for general intelligence factor - is still seen as the dominant variable in determining the intellectual performance of our brain. (G was first coined, in 1904, by the psychometrician Charles Spearman, who noticed that the grades of young kids were correlated across seemingly unrelated subjects.) The first thing to say about g is that it's an incredibly robust statistical phenomenon. This means that the same person will get a similar score on an IQ test at the age of 12, 20, and 50. Furthermore, his score will correlate nicely with his academic performance, at least in certain subjects. For instance, a 2007 study by psychologists at the University of Edinburgh found that general intelligence accounted for 58.6% of the individual variance in math performance, 48% of the variance in english, and 18.1% of the variance in art and design.

Of course, that still leaves a lot of variance unaccounted for, even in those academic subjects, like math, that are supposed to depend on the very mental skills measured by IQ tests. This helps explain why Lewis Terman, the inventor of the Stanford-Binet IQ test, eventually became frustrated with his measurement. Terman spent decades following a large sample of "gifted" students, searching for evidence that his test of intelligence was linked to real world success. While the most accomplished men did have slightly higher scores, Terman eventually concluded that other factors played an even more important role. He argued that one of the most fundamental tasks of modern psychology was to figure out why general intelligence is not a more important part of achievement: "Why this is so, and what circumstances affect the fruition of human talent, are questions of such transcendent importance that they should be investigated by every method that promises the slightest reduction of our present ignorance."

In order to understand the limitations of general intelligence, at least as presently defined, it's important to delve into one of the of the great themes of modern psychology, which is the essential role of the unconscious. While Freud associated the unconscious with the unspeakable urges of the id, we now know that our mental underworld is actually a remarkable information processing device, which helps us make sense of reality.

This has led to the dual process model of cognition, in which the mind is divided into two general modes. There is Type 1 thinking, which is largely unconscious, automatic, contextual, emotional and speedy; it turns out that most of our behavior is shaped by these inarticulate thoughts. (Consider, for instance, what happens when you brake for a yellow light, or order a dish on a menu as soon as you see it, or have an "intuition" about how to approach a problem.) And then there is Type 2 thinking, which is deliberate, explicit, effortful and intentional. (Imagine an amateur chess player, contemplating the implications of each potential move.) Needless to say, intelligence tests excel at measuring Type 2 thought processes, which is why the standard IQ test largely relies on abstract puzzles and math problems, and correlates with working memory performance.

The end result is a growing contradiction between how we define intelligence - it's all about explicit thought and g - and how we conceptualize cognition, which is inextricably bound up with Type 1 processes. (In other words, we currently measure intelligence by excluded the vast majority of the information processing taking place inside our head.)

Furthermore, this obsession with individual variation as measured by g has meant there's been virtually no investigation of individual variation when it comes to the output of the unconscious, or the speed/efficiency of Type 1 thinking. We've assumed that the subterranean brain - this primal, Pleistocene supercomputer - is virtually uniform and universal, and runs the same stupid software programs in everyone. Here's a sample excerpt from a recent review on Type 1 thinking: "Continuous individual differences [in unconscious mental processes] are few. The individual differences that do exist largely reflect damage to cognitive modules that result in very discontinuous cognitive dysfunction such as autism or the agnosias and alexias." In other words, the variance that matters exists in Type 2 thinking.

In recent years, however, this assumption has begun to fall apart. There's a growing body of evidence that reliable differences exist in Type 1 thinking, and that these differences have consequences. This helps explain why even the most mundane features of Type 1 thinking - such as serial reaction time - significantly correlate with math and verbal scores on the ACT. Other studies have found that performance on a variety of implicit learning tasks - the kind of learning that takes place in Type 1 - were significantly associated with academic performance, even when "psychometric intelligence," or g, was controlled for. In other words, not every unconscious works the same way.

This view of Type 1 thinking as an individual "ability" with meaningful individual differences is the subject of an important new study, "Implicit learning as an ability," in Cognition led by Scott Barry Kaufman, a psychologist at NYU. The scientists began by measuring implicit learning performance on 153 adolescent students. Sure enough, they found reliable differences between subjects, so that some students were consistently better at "automatically and implicitly detecting complex and noisy regularities in the environment".

Here's where the data gets really interesting: These individual differences in unconscious processing correlated with academic performance on a wide range of subjects, from foreign language to math. In other words, students who did better on the seemingly mindless implicit learning task were also better at conjugating French verbs, even when controlling for the effect of "psychometric intelligence". This clearly demonstrates that much of our intellectual variation has nothing do with the intellectual skills we measure and valorize. Instead, our intelligence is deeply influenced by all sorts of subliminal talents that we can't control, influence or directly access. Here's the conclusion of the Kaufman paper:

The pattern of variables that are and are not related to implicit learning is suggestive of conclusions about the structure of human information processing, consistent with the idea that there are two relatively independent systems by which individuals analyze and learn about regularities in their experience. Further, these results suggest that the investigation of individual differences in implicit cognition can increase our understanding of human intelligence, personality, skill acquisition, and language acquisition specifically, as well as human complex cognition more generally.

Needless to say, these results raise plenty of important questions. The one I'm most interested in is whether or not these Type 1/implicit learning skills can be improved over time. While numerous studies have demonstrated that experience can improve the performance of the unconscious in extremely specific contexts - a Nascar driver will have better driving reflexes - there has been little research into the malleability of the overarching Type 1 system. If these unconscious/implicit learning skills turn out to be teachable, then we suddenly have an entirely new way to educating the brain and improving cognitive performance.


More like this

Whether we're talking about type 1 (i.e., "g") or type 2 intelligence, and which regions of the brain are responsible for which, there's still the question of what that intelligence actually is. I understand that there's been a lot of debate over this question.

But isn't it just pattern ability? That is, the ability to identify, store, encode, utilize, and create patterns accurately and efficiently? What else is there to intelligence?

This line of research is very intriguing and exciting, given that g clearly does not account for all of the variance in cognitive tasks. However, I (of course) would really like to see the idea that g has little relationship to implicit learning replicated. G has been found to positively correlate virtually every cognitive performance tasks tested in much larger samples (see Robert Plomin's work at King's College London), and even to other physiological health markers such as height and semen quality (no joke). It seems counter-intuitive that it wouldn't be here as well.

That said, the underlying ideas here are very fascinating. I agree with Kaufman et al. that individual differences in the implicit learning system are meaningful and should be taken seriously, and the results with the personality measures, in particular, really intrigue me. I hope to see more work in this vein! (And great post, by the way.)


As a school psychologist, having administered thousands of IQ and achievement assessments to children, I'm fascinated by this post. The vast majority of assessments were to determine if student had a learning disability, which is a significant discrepancy between intellectual ability and achievement. I've been discouraged by lack of explanation for causal factors when student has average or above avg IQ abilities but below avg achievement, when not explained by other factors, such as motivation, prior learning, environmental factors, etc. Perhaps deficit(s) or weakness(s) in type 1 cognition is possible causal explanation?

As the brain is exposed to more fight or flight snap decisions or choices with higher risk associated, the "smarter" we become, or so the early data may indicate. Considering what types of cognitive processes successfully boost survival rates through the generations, it would not be too surprising to find adaptability is at the fulcrum of our more later more specialized cognitive advances. This doesn't say much yet on how that heightened stimulation might complicate or trade off with the emotional high cost associated with these elevations in stress, certainly there is data surfacing about the physical limitations to overstressed environments, but it would suggest a structured level of empowered decision making tasks could help develop core cognitive ability around road blocks in traditional learning, disabilities and alternate specialized preference. That would be enriching indeed.

As a parent, I would be interested in research that explores how to improve type 1 cognition. Is it a matter of playing certain games with my son? Talking to him in a certain way? Singing more? etc etc

Matt -

I agree that this innate intelligence is essentially pattern recognition. There are many qualities to this process though, and there in may lie the differences between individuals.

For instance, how quickly do "patterns" emerge, i.e. how many experiences of a pattern are required for recognition? It may seem obvious that this depends on the pattern (a complex pattern will jump out quickly, while more subtle patterns take time to form), but perhaps more intelligent subconscious require fewer patterns to put the pieces together.

Another quality may be the breadth at which seemingly incongruous patterns can manipulated and coerced into larger patterns, i.e. the ability to apply a concept from one domain into another.

There are undoubtedly more such qualities of the subconscious. I think improving the innate intelligence that Jonah talks about would be akin to optimizing these emergent properties.

This is a fascinating post. My personal intuition is that type 1/implicit learning is malleable, though I have no convictions on whether or not it can be improved. These questions when asked in conjunction with questions concerning what the internet is doing to our brains (or not)would be all that more interesting.

By Ryan Sittler (not verified) on 28 Jun 2010 #permalink

I would be very interested in the effects of video games on implicit learning. By doing "knee-jerk" tasks over and over, do we improve it?

It also ties nicely into the Tim Howard factor from your previous post.

"- consistent with the idea that there are two relatively independent systems by which individuals analyze and learn about regularities in their experience."
I expect that these "relatively" independent systems will be discovered to be more relatively dependent than the alternative.
We may think we're using the type2 or rational brain most often when taking an IQ test, but this is mainly because that's the part of the "thinking" we are conscious of. But the intuitive part of the type1 mind is also thought to govern the extent or level to which the type2 brain can deal with the abstractions that it supposes it has originated.
Note that you have just referred to your personal intuition as doing just that. Your unconscious brain has dictated your thoughts for you to write - which to that the end becomes a conscious process.

What surprised me in this post (and the comments that follow) was the degree of doubt that unconscious intelligence is behind measurable, explicit intelligence. I'm not a psychologist, but have read a lot on the subject, and my understanding is that much of our conscious decision making (and reasoning) has been shown to be a function of our brain creating a rationale for what we have already concluded unconsciously. Those implicit reactions are many times faster than conscious we tend to respond before we are aware of why we are doing so.

Perhaps those with strong "g"/explicit intelligence are simply better at deciphering and communicating the reasoning behind their more automatic/implicit responses to what they've learned. Or perhaps I a naive amateur who doesn't know what she's talking about!?

By Laurie Rippon (not verified) on 28 Jun 2010 #permalink

Fascinating stuff. I wonder to what extent type 1 and type 2 thinking can be identified with the distict left and right hemisphere activities as they seem quite closely correlated as I understand the processes of the hemispheres to be (and here I admit my conception of the difference is largely drawn from Jill Sobule's excellent TED talk).

I also found myself considering areas of learning at school that I excelled at or sucked at. I reckon that in order to determine ones relative intelligence levels in type 1 and type 2 thinking we should control for such hard-to-measure variables as boredom, ennui, overall general interest in a particular subject, how engaging a teacher is with the individual, how well explained a subject is, and a child's general level of creativity... surely for everyone learning calculus is easier if you care about the problem to which calculus is the solution.

In other words I get the distict impression - if I have understood correctly - that measurements of type 1 thinking would have been a poor indicator of academic achievement in my case.

The dual process model of cognition is very similar to the architectures used by engineers to design robots. With robots, you have reactive and deliberative behaviors, which correspond to the Type 1 and Type 2 models respectively. A reactive/Type 1 behavior is what a vacuum robot does when it hits a wall. The robot will turn around and keep cleaning till it hits another wall. A deliberative/Type 2 behavior would be similar to what the entrants in the DARPA grand challenge did. The robots were tasked to navigate to a destination through an unknown area. It would be interesting to do a deeper comparison of the two fields.

I meant Jill Bolte Taylor. Jill Sobule is in my other hemisphere.


Perhaps relevant: Many years ago CCL (The Center for Creative Leadership) took a look at why Ivy Leaguers weren't well represented among the "best and brightest" of corporate America.

Their down and dirty, as I recall, was a division between "academic intelligence" and "street smarts." The folks running corporations were--not surprisingly perhaps--high on street smarts.

And don't we all know "bright" folks who have difficult navigating their way out of paper bags? And folks who didn't do so well in school, but are doing just fine, thank you very much, in the marketplace?

By Cat Peterson (not verified) on 29 Jun 2010 #permalink

Hi, I have been pondering this type of question for some time now as it has relevance to decision making and human performance at work .. and I propose that there are actually 4 decision types: AUTOMATIC which are the things we do with zero thought possibly including things that have been learnt to such a degree they can be performed perfectly well without consciousness - like kicking a ball/knitting; then there is CONTINUOUS which are the very low level effort in line with Tier 1; VARIABLE which is tier 2; and finally REGULATORY which are the things that are conceptual or abstracted that require a combination of logic and intuition. You can catagorise all processes within one these four types and for each of these tiers you can analyse and outline the characteristics that define the attributes for the tier as they are largely the same- like feedback mechanism, quality or error control, decision method, etc. If you take a laptop you have Automatic functions like the operating system, you have Continuous like the mouse or keyboard, you have Variable such as the applications you are running, and you have Regulatory for back-ups/defrags/updates to virus apps etc. You can take many things and apply this 4 tier approach (cars, companies, schools, web sites, etc). This also maps onto the way our brain is seen to develop and grow. I have also looked at how processes or tasks move between the tiers as they do - riding a bike was really difficult at first but by the time you are 20 its very easy so its moved from Tier 2 to Tier 1 possibly down to my Automatic. In fact I suggest that this is what the brain is designed to do - to get more efficient and perform tasks with less effort or brain power. I am in the middle of published my complete work on this subject and have called it the Theory of Human Efficiency (T.H.E.) so any feedback would be welcome on this. Thanks - Terry

By terry wilcox (not verified) on 29 Jun 2010 #permalink

I would be very interested in the effects of video games on implicit learning. By doing "knee-jerk" tasks over and over, do we improve it?

My youngest son is mildly autistic and even now, at nearly age eight, has difficulty making his speech understood by most people. His socialization was stunted and he was timid to expose his disability. As recently as a year ago he was basically unintelligible to anyone but immediate family.

Wii and Nintendo DS starting a few months before his sixth birthday made a huge difference - literally within a few months. Finding something he was good at and got him social recognition was a big part of it of course. But the games did more than that. We were stunned at how quickly (just a few months) his pattern/problem solving went through the roof. It goes well beyond reflexes - he mastered basic logistics in no time and now plays the xBox version of Civilization.

He regularly sees excellent counselors and his school works with him as well. Everything comes together and reinforces his progress.

But it was video games that made the most dramatic and immediate difference. It helped his organize his brain, and was a positive feed forward to his confidence and social competence.

Related comment: we've talked for years about left/right brain and the successivly more complex regions of specialization. But I'm particularly interested in how the glial region brings it all together. I suspect much of what we think of as intuition and creative innovation is actual the meta influence of the glial reaction.

I'm a layman and no doubt doing terrible things to the neuroscience field with my oversimplification. But I can't help but feel we're going to uncover much that is currently poorly understood with the recent discovery that neuroglia is far more than just glue...

How much of the left-over, non-IQ variance (for math ability, for example) is explained by type I thinking?

Somehow I doubt very much.

This is merely an impression, but it might be interesting. I was extremely good at IQ and other standardized tests (usually above the 99th %tile in the verbal parts, about the 96h or 97th %tile in the math) (I'm male). and I was also extremely fast, that is, I almost always finished each section ahead of time. I am quite sure that my first time through each part depended heavily on nonlinear, 'intuitive' thinking, which often involved psyching out the question rather than thinking it through systematically, particularly in the easier parts. If I checked my answers, I slowed down and became much more systematic.

I suspect, as the field evolves, the metaphors for how we think and respond are going to keep changing. This is good.

In the 19th century everything was a clock-work, and the universe was defined as a big gear-box, because that was the cutting edge technology of the time. Our brains were seen to be basically like a swiss watch, only with thousands of more tiny gears.

Nowadays, computers are the cutting edge technology, and so we are describing our brains and bodies as being computers that operate on programs and sub-routines, hardware and software. Our brains are described as if they were desktop computers, only with lots more sub-routines coded in.

I want to caution about getting over-attached to our metaphors. We have to be open to the idea that maybe the brain is not like a desktop computer. Maybe the way connections are set up, changed, maintained and re-foremd in the brain does NOT work like machine binary code, with higher-level languages like C++ code piled on top of it.

This is the problem with all types of modelling. While modelling can help us understand processes over time, there's always the danger that we start becoming so enthralled by the pretty model that we move further away from from the reality that the model was developed to represent.

Just my two-cents.

You may be interested in a recent (2009) article co-authored by Daniel Kahneman: Conditions for Intuitive Expertise: A Failure to Disagree.

One of the things he examines in this article is what it takes to be a better intuitive thinker, with the idea being that certain environmental factors can help us get better at this "Type 1" thinking. It's a marvelous article.

Here's a link to the abstract:

Aren't we really talking about instinct? (Type 1) vs. voluntary or directed thinking? (Type 2) and are the two types (as defined here) actually separate, or are they integrated? Doesn't voluntary cognition become the arbitrator of "what to do about" instinct? Why would anyone think that instinct can be taught? It's there; automatic, involuntary, expressed as emotions and intuition. Language allows us to express and work with instinct in useful and socially acceptable ways. When we DON'T teach language skills we're left with emotionally confused children - they can't make sense of their feelings or the environment. Let's get rid of this "teaching children basic skills is cruel and unusual punishment" thing once and for all. Humans have an instinct to learn, which modern educators have managed to destroy by denying instinct. We learn by copying - imitating people who know how to do something, whether it's swinging a hammer or playing the piano. Very little "learning" is at the abstract level.

Video games exploit instinct, but is that what we really want? Kids who know how to kill, maim, torture, and enjoy the destruction of other human beings?

I'm curious to know whether Type 1 thinking may somehow correlate with the though process that underpins Design Thinking (also known as Integrative Thinking)? Popularized by the likes of IDEO and the Standford, this new school of thought, if it can be called such, is said to provide a problem-solving alternative to the rational and logical decision-making processâan alternative more heavily concerned with intuition, empathy, and creativity. The gist of it is basically that designers (Type 1?) make better decisions (in some areas) than, say, MBAs (Type 2?) because they are more "open minded". I've yet to find any factual evidence behind this bold claim, but maybe the Type 1 and Type 2 delineation can be provide a starting point?

Thanks for this thought-provoking post.
I read the paper, which is beautifully clear and well-argued, but I was concerned at the relatively low reliability (.44) of the implicit learning measure - esp. as this was a split half reliability. For research on individual differencs in implicit learning to move forward, we'll need measures that show good test-retest reliability - ie we need to be confident we are measuring a stable individual difference. That will be challenging because of potential interference between test occasions.
The reported correlations with verbal reasoning etc are pretty small (around .2) and don't account for much variance - question is, whether they would look better if we had a more reliable measure?

By Dorothy Bishop (not verified) on 10 Jul 2010 #permalink

знаеÑе о Ñем Ñ Ð¼ÐµÑÑаÑ?
ÑÑоб в один пÑекÑаÑнÑй Ð´ÐµÐ½Ñ ÑобÑалиÑÑ Ð²Ñ Ð²Ñе на болÑÑом майдане
вÑе гоÑÐ»Ð¾Ð¿Ð°Ð½Ñ ÐºÑикÑÐ½Ñ ÑеÑевÑе ÑомÑÑки боÑÑÑ Ñ ÐºÐ¾ÑÑпÑией вÑÑÑÑпаÑÑие за влаÑÑÑ Ð½Ð°Ñода завиÑÑливÑе ÐºÐ¾Ð·Ð»Ñ ÑÑиÑаÑÑие денÑги в ÑÑжом коÑмане манипÑлиÑÑемÑе лоÑи и неÑдаÑники вÑÐµÑ Ð¼Ð°ÑÑей обвинÑÑÑие в ÑÑом кого Ñгодно ÑолÑко не ÑебÑ
ÑобÑалиÑÑ Ñо вÑеми ваÑими знаменами
Ñ Ð¿ÐµÐ½Ð¾Ð¹ Ñ ÑÐ°Ñ Ð¸ кÑиками Ñиновников на ÑонаÑи!!! Ð½ÐµÑ ÐºÐ¾ÑÑÑпÑии! влаÑÑÑ -наÑодÑ!
ÑÑоб ÑобÑалиÑÑ Ð²Ñе
ÑÑоб ни один ÑÑка дома не оÑÑалÑÑ
главное ÑÑоб вÑе вÑлезли
Ð¶Ð°Ð»Ñ ÑолÑко плоÑади Ñакой Ð½ÐµÑ Ð²
Ð²Ð¾Ñ Ð² киÑае ÑоÑоÑо бÑло
ТÑÐ½Ñ Ð°Ð½Ñ Ð¼ÑÑÐ½Ñ Ð²Ð¼ÐµÑÐ°ÐµÑ 1 миллион Ñеловек

и Ð²Ð¾Ñ ÐºÐ¾Ð³Ð´Ð° вÑе Ð±Ñ Ð²Ñ Ð²Ñлезли
вÑÑла Ð±Ñ ÑÐ°Ð½ÐºÐ¾Ð²Ð°Ñ Ð°ÑÐ¼Ð¸Ñ Ð¸ вÑÑ ÑволоÑÑ Ð²Ñе говно наÑии намаÑало Ð±Ñ Ð½Ð° гÑÑениÑÑ
Ñжгло Ð±Ñ Ð²Ñе каленÑм железом
и Ð²Ð¾Ñ Ñогда как поÑле ТÑнÑÐ°Ð½Ñ Ð¼ÑÐ½Ñ Ð² ÐиÑае Ñ Ð½Ð°Ñ Ð±Ñ Ñоже 20-30 Ð»ÐµÑ Ð¿Ð¾Ð´ÑÑд бÑл Ð±Ñ ÑкономиÑеÑкий ÑоÑÑ Ð¿Ð¾ 10 пÑоÑенÑов в год

а ÑеÑез 20-30 Ð»ÐµÑ Ð²ÑÑ Ð¿ÑоÑедÑÑÑ Ð¾Ð¿ÑÑÑ Ð¿Ð¾Ð²ÑоÑиÑÑ Ð¿Ð¾ÑÐ¾Ð¼Ñ ÑÑо опÑÑÑ Ð½Ð°ÑодиÑÑÑÑ Ð½Ð¾Ð²Ð³Ð¾Ðµ поколение майданÑиков неÑдаÑников лоÑов и гоÑлопанов

ÑÑ Ðº наÑÐ¾Ð´Ñ Ð¸Ð¼ÐµÐµÑÑ Ñакое же оÑноÑение как кал к ÑелÑ
а вÑÑкое Ñело пеÑиодиÑеÑки должно ÑамооÑиÑаÑÑÑÑ
пÑÑем ÑÑнÑанбмÑнÑ
инаÑе, еÑли либеÑалÑниÑаÑÑ ÐºÐ°Ð» майданÑÑÑÑ Ð¸ впÑÐ¼Ñ Ð¼Ð¾Ð¶ÐµÑ Ð¿Ð¾Ð¿Ð°ÑÑÑ Ð² мозг
как ÑÑо бÑло на ÑкÑаине поÑле майдана
и за 5 Ð»ÐµÑ ÑÑаÑна бÑла оÑбÑоÑена на 10 Ð»ÐµÑ Ð½Ð°Ð·Ð°Ð´

или как Ñ Ð½Ð°Ñ Ð² ÑеволÑÑÐ¸Ñ Ð¿Ð¾Ñле вÑÑÑÑела авÑоÑÑ
когда ÑÑаÑна 10 миллионов Ñеловек поÑеÑÑла из за влаÑÑи гоÑлопанов
ÑоÑоÑо ÑоваÑÐ¸Ñ ÑÑалин ÑÑÐ¸Ñ Ð¿ÐµÑемненÑÑÑ ÑеволÑÑионеÑов и гоÑлопанов ÑеволÑÑии и ленинÑкÑÑ Ð³Ð²Ð°ÑÐ´Ð¸Ñ Ð¾ÑпÑавил ÑабоÑаÑÑ Ð½Ð° ÑевеÑ

мой ÑиновниÑÑий жÑÑнал