Kovacs & Conway’s (2016) Process Overlap Theory (POT) is the latest general, unified theory of intelligence based on the current body of evidence, with profound implications for cognitive training and intelligence augmentation.
.
It is a good starting point for understanding the nature of intelligence and the human mind more generally, and is worth reading digested content here to get a grasp of the basics.
.
 |
.
Since its recent entry on the scene, Process Overlap Theory has already created waves in the field of intelligence, shaking more conventional views about the nature of intelligence. It is a theory of g that is endorsed by Prof. Kevin McGrew, Director of the Applied Institute of Psychometrics – the scientist behind the highly influential CHC Theory of cognitive ability you may be familiar with from my own eBooks, web content and blog posts.
.
The theory goes beyond mere psychometrics and statistics (correlations of test scores and the like): it offers a unified explanation of the underlying cognitive processes and neural mechanisms.
.
Let’s introduce the theory by reviewing some basic concepts.
.
Where does the idea of g come from?
We know where the concept of general intelligence came from. It comes from the fact that people who perform above average on one kind of cognitive test (e.g., vocabulary) tend to perform above average on other kinds of cognitive tests as well (e.g., mental rotation or matrices), and we say that g represents an underlying general ability that explains the positive correlations.
.
But is g something real in our brains and cognitive processes? Does it actually explain anything?
.
POT Premise 1: g is a statistical fiction – not something real in brains
.
It turns out that g as a mathematical entity is a necessary (logical) consequence of positive correlations between tests. Based on the laws of mathematics it is always possible to extract a single ‘general factor’ when there are only positive correlations. But that doesn’t mean such a factor exists in the real world or that it can explain why the correlations exist in the first place.
Nisbett et al. (2012) in their review for the American Psychological Association, point out, such studies have been inconsistent:
“Patterns of activation in response to various fluid reasoning tasks are diverse, and brain regions activated in response to ostensibly similar types of reasoning (inductive, deductive) appear to be closely associated with task content and context. The evidence is not consistent with the view that there is a unitary reasoning neural substrate.” (p. 145)
Haier et al. (2009) also conclude after similar inconsistent results that “identifying a ‘neuro-g’ will be difficult”.
.
Fact 1: To date no one has succeeded in measuring or describing anything in the brain that can be identified with g (general intelligence) – whether as, neural efficiency, a ‘general intelligence’ network, network tuning, amount of brain entropy and so on.
.
Let’s look more closely at what g actually is if it’s not something in brains.
.
Statistics 101: Explanatory power of g & multiple intelligences
.
People different on how well they score on any given cognitive test such as the Raven Advanced Progressive Matrices. The range or ‘spread’ of different scores among all test-takers is called the variance of test scores in the population. When the variance is high, there is a wide range of scores, from very low to very high. Variance is a measure of populations not individual test takers.
.
When there is a high positive correlation between scores on different types of cognitive tests (Raven’s, vocabulary, mental rotation, reaction time, etc) statisticians say that a general factor (g) ‘explains’ a lot of the variance in test scores in the population of test takers: it has high explanatory power. When there is a low (average) correlation between different test scores, statisticians say that more of the variance is explained by test-specific abilities, and less by a general factor (g): it has low explanatory power.
.
So strictly speaking, g is a statistical concept that is used to explain variations in populations; it does not make sense to use g to explain anything going on in individuals. As Kovacs & Conway put it:
.
“—the last survivor of a meteor collision with Earth would still have cognitive abilities and mental limitations but would not have g.” (p. 153)
.
But let us suppose that g can be interpreted somehow in terms of different underlying cognitive processes or neural mechanisms in populations. .In this case, you might expect that populations with high IQs (e.g. Harvard graduates) would have more g than populations with a low IQs (e.g. a low ranking gang members in and out of the prison system).
.
But it turns out the opposite is the case!
.
Fact 2: General intelligence (g) explains more of the differences in mental ability at lower levels of ability than at higher levels of ability. The higher the IQ, the less g predicts IQ scores, and the more sub-test specific cognitive skill-sets explain the differences in overall IQ scores. As cognitive ability increases (as measured by IQ tests), the less explanatory power g has. Higher ability groups in fact have less g – not more g!
.
Similarly –
.
Fact 3: The higher a nation scores on IQ tests (e.g. South Korea), the less g explains that nation’s test scores and the more test-specific skills explains the scores.
.
And what’s more –
.
Fact 4: Gains in IQ from one generation to the next (the well-known Flynn effect) are accompanied by decrease in how useful the general factor ‘g’ is for predicting IQ test scores.
.
So to summarize these facts:
.
General Fact: The explanatory power of g – how useful it is in explaining cognitive performance – varies as a function of cognitive ability level. The higher the cognitive ability (as measured by IQ tests), the worse the explanatory power of g – and the better the explanatory power of test-specific multiple intelligences.
.
The research behind all of this can be found clearly laid out in the Kovacs & Conway’s (2016) article (ref).
.
So if g is only a statistical abstraction and not real in terms of psychological processes and neural mechanisms, and even as a statistical entity has less and less explanatory power in populations higher levels of cognitive ability, what is real and what is more explanatory?
.
POT Premise 2: What IS real are general executive processes
.
The POT theory states that what is real in the brain are (a) domain general executive processes and (b) domain specific processes such as visuospatial or verbal processes- shown by the black dots and ‘V’s and ‘S’s in the POT model shown here. These can all be located in actual brains! Only two types of domain specific process are shown in the model below – but there are others including quantitative/numerical.
.
 |
.
The Process Overlap Theory of General Intelligence (g)
The g factor in this model is just a statistical fiction – extracted at the mathematical-statistical level as a kind of summary of the underlying correlations between the cognitive tests (such as matrices, number progression, vocabulary). These tests are represented in the model by the colored boxes – three tests for for each broad IQ ability shown – fluid intelligence (Gf), crystallized intelligence (Gc), and visuospatial intelligence (Gv). It is the overlap of the same executive processes (black dots) that actually explains the positive correlations between all the cognitive tests – not the mathematical general factor g. Hence ‘process overlap theory’. Different tests within the same broad ability measure different subsets of these executive processes, and their interactions with domain-specific processes – as shown by the colored ovals.
.
What are the domain-general executive processes?
These are flexible, general purpose executive functions such as working memory, inhibition and attention focus and flexibility (trained in IQ Mindware apps), that can be applied to any type of domain – whether verbal, visuospatial, numerical, etc. That is, they are ‘domain general’. Here is one representation of candidtate executive processes (‘workspace’ = working memory workspace).
.
 |
.
Together these processes act like a bottleneck on cognition. They are limited capacity in terms of the information they can process at any one time. It is because of their limited capacity we cannot ‘think through’ two problems at the same exact time – our focus of attention can only be on one problem at a time, even when we multitask. These attention-based executive processes are located in well-known executive control brain networks in pre-frontal and parietal regions in the brain that I have discussed in my eBooks. The locations of the hubs of these neural networks are shown here:
.
 |
.
POT Premise 3: Executive Processes = Fluid Intelligence (Gf)
.
POT claims that these executive processes together are the same as fluid intelligence (Gf) when they are applied to solving complex and novel cognitive challenges, such as problem solving, reasoning, comprehending and skill-learning at the highest levels of abstraction. This is also the most general-purpose, adaptive level of cognition..Subsets of the different executive processes are measured by fluid intelligence tests such as Raven’s matrices, number series, or concept formation tests – as shown in the POT model above.
.
Fluid intelligence (Gf) definition: The ability to solve problems in novel, complex situations, regardless of previously acquired knowledge. Fluid intelligence is the ability to reason, problem solve, and to see patterns or relations while thinking. It includes both inductive and deductive logical reasoning. It involves being able to figure out the abstract relations underlying analogies.
.
The POT identification of Gf with the executive processes underlying the correlations between all cognitive tests is consistent with these two facts – the last we will consider today.
.
Fact 5: g is virtually identical with Gf at a statistical factor level. A theory of intelligence must account for the central role of fluid abilities in g.
.
Fact 6: More complex tests (e.g. matrices, comprehension) are better explained by g (have higher positive correlations among themselves) than less complex tests (reaction time, vocabulary).
.
Fact 6 is explained according to Process Overlap Theory by virtue of the fact that more complex tests have more overlapping executive processes involved than less complex tests.