The unsolved problems of…

Speaking of the brain, Ralph Adolphs published in 2015 one short and thought-provoking article addressing the unsolved problems of neuroscience. He identified three meta-questions:

1. a. What counts as understanding the brain?

2. b. How can a brain be built?

3. c. What are the different ways of understanding the brain?

He highlighted that the most important unsolved problem is “how the brain generates the mind”. This can be translated into The Brain Connection (TBC) view (Colom, Reference Colom2015): the biggest unsolved problem is how the brain generates intelligence from the genotype and from the environmental relevant ingredients.

Nevertheless, perhaps the most suggestive issue discussed by Adolphs relates to the distinction among different types of problems: (four) problems that are solved (or soon will be), (eight) problems that we should be able to solve in the next fifty years, (five) problems that we should be able to solve but who knows when, and (three) problems we may never solve (see his Box 2).

We already know how single neurons compute and what is the connectome of small nervous systems, but we still lack reliable knowledge regarding how neuronal circuits compute or how brains produce decisions. Consistent with TBC view, Adolphs thinks we should be able to solve two key problems:

1. a. What is the complete connectome of the human brain?

2. b. How could we make everybody’s brain function best?

In this respect, on his monumental book about intelligence research, Hunt (Reference Hunt2011a) admits the crucial role of the brain for understanding human intelligence. Indeed, everything related with intelligent behavior is supported by actions taking place in our brains: “if we knew the nature of every connection between the approx. five billion neurons in a person’s brain, and if we knew the algorithms the brain uses to activate and alter these connections, we would know everything there is to know about that person’s cognition” (Hunt, Reference Hunt2011a).

On his recent inspiring book about the neuroscience of intelligence, Richard J. Haier acknowledges that, at the end of the day, scientific research is oriented towards the key goal of finding efficient ways for enhancing the intelligence level of the population (Haier, Reference Haier2016). But for achieving this ultimate (and desirable) purpose, we must understand how our brains produce our intelligence (TBC again).

Let me close this section noting that Ralph Adolphs considers that the ‘mind’ includes cognition, computation, information processing, thinking, and reasoning. Interestingly, all these features are comprised by the shared definition of intelligence: “a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly, and learn from experience” (Gottfredson, Reference Gottfredson1997). Adolphs himself goes further by assuming that all these features are somehow related. Intelligence researchers know very well what this fact involves. I will return to this key issue below.

Intelligence according to the American Psychological Association (1996–2012)

The APA reports on intelligence published fifteen years apart are useful here because they identified ‘knowns and unknowns’ within the intelligence research area. These were the main points noted by Neisser et al. (Reference Neisser, Boodoo, Bouchard, Boykin, Brody, Ceci and Urbina1996) after their exhaustive revision of the available evidence:

• - Research has identified correlations of cognitive (information-processing) and brain measures with psychometric intelligence. However, we do not know why these factors are related.

• - Genetic differences contribute to intelligence differences. However, we do not know the mechanics linking genes and behavior.

• - The impact of genetic differences increases across the life span. However, we do not know why this happens.

• - We know the environment is important, but we still lack solid knowledge regarding the specific factors contributing to intelligence differences.

• - Intelligence test scores have meaningfully increased in the last decades (Flynn effect). However, scientists still argue about the causes behind this secular trend.

• - Identified average population differences (sex, race) do not result from biases related with test construction or test administration.

Nisbett et al.’s (Reference Nisbett, Aronson, Blair, Dickens, Flynn, Halpern and Turkheimer2012) updated this report and five were the most relevant points:

• - There is a strong correlation between intelligence and working memory capacity, but we still don’t know why.

• - Secular gains in intelligence (Flynn effect) are environmentally caused and this fact is consistent with the strong genetic influence over intelligence differences within generations.

• - The general factor of intelligence (g) is unitary at the psychometric level, but there are models suggesting that it is not unitary at the cognitive and brain levels. The ‘Process Overlap Theory’ proposed by Kovacs and Conway (Reference Kovacs and Conway2016) is a recent example (see Colom, Chuderski, & Santarnecchi, Reference Colom, Chuderski and Santarnecchi2016; for a critical comment).

• - Presumed non-cognitive psychological factors such as ‘self-control’ might be relevant for understanding intellectual performance differences.

• - Finally, ‘stress’ may have some impact over the observed group differences in intelligence.

Both reports acknowledge how research based on psychometric methods greatly contributed to our impressive current knowledge, and they also invite to increase shared and sustained efforts for going one step further avoiding crude political assertions. The XXI Century must move beyond old-fashioned and outdated arguments regarding this psychological factor. Thus, for instance, we already know test scores do reflect intelligence and saying that these scores are not socially important ignores the available scientific facts. Furthermore, these facts dramatically fail to support the popular perspective that intelligence can be described by a set of independent cognitive abilities. We know these abilities are related to a substantial extent (Hunt, Reference Hunt2011a). Revisiting these core issues again and again is not really smart or wise.

Interlude

In this interlude, I want to highlight, before moving ahead, four key issues related with the presentations of this international seminar.

First, as said before, the psychometric approach has been very useful for expanding our understanding of the intelligence construct. Furthermore, millions of intelligence tests are administered annually worldwide because they are appreciated as reliable and valid assessment devices (Detterman, Reference Detterman2014). These tests measure cognitive abilities unique to the testing situation (testing skills), cognitive abilities relevant for our society (reasoning), and cognitive abilities common to humankind (spatial-visual reasoning and memory) (Hunt, Reference Hunt2011a). Detterman’s presentation discussed evidence leading to the unappreciated finding that students’ cognitive abilities measured by these standardized intelligence tests largely surpass traditionally underscored factors, such as teachers and schools, for predicting learning outcomes.

Classic standardized testing will be with us for a while, but we must move beyond these classic approaches for monitoring behavior outside standardized testing. We must invest our resources for moving from conventional testing procedures to the rigorous measurement of intelligent behavior in settings as close as possible to everyday life (Hunt, Reference Hunt, Sternberg and Barry Kaufman2011b). In this regard, Quiroga’s proposal for measuring intelligence using video games seems highly relevant.

Second, general intelligence (g) is mainly related with two information-processing cognitive abilities, namely, working memory capacity (WMC) and speed of information processing. Colom et al. (Reference Colom, Chuderski and Santarnecchi2016) revised this issue at length, concluding that the strong relationship between g and WMC can be explained by a common capacity (abstract working memory) or the ability to construct and maintain arbitrary bindings (relational integration). Basic executive control processes, such as attention, interference resolution, and inhibition, are much less relevant in this regard. Furthermore, processing speed seems to play some role only when children or older adults are considered (Tourva, Spanoudis, & Demetriou, Reference Tourva, Spanoudis and Demetriou2016).

Third, more than four centuries ago, Huarte de San Juan (Reference Huarte de San Juan1575) acknowledged that intelligence must have one strong biological substrate. We now know this is the case, but we are also still looking for answers. Nevertheless, we have learned many interesting things since Huarte’s times. Stuart Ritchie, Emiliano Santarnecchi, Adam Chuderski, and Norbert Jausovec discussed several of these things on their presentations, mainly focused on functional data.

The available neuroscience evidence is generally consistent with the parieto-frontal integration theory (P-FIT) of intelligence (Jung & Haier, Reference Jung and Haier2007) meaning that individual differences in frontal and parietal regions, along with their communication, are especially relevant for supporting intelligence differences (Barbey et al., Reference Barbey, Colom, Solomon, Krueger, Forbes and Grafman2012; Colom et al., Reference Colom, Haier, Head, Álvarez-Linera, Quiroga, Shih and Jung2009; Gläscher et al., Reference Gläscher, Rudrauf, Colom, Paul, Tranel, Damasio and Adolphs2010; Pineda-Pardo, Martínez, Román, & Colom, Reference Pineda-Pardo, Martínez, Román and Colom2016). However, available findings also show that different brain networks might be relevant for different individuals (Martínez et al., Reference Martínez, Madsen, Joshi, Shantanu, Román, Villalón-Reina and Colom2015). In this regard, the recent meta-analysis by Basten, Hilger, and Fiebach (Reference Basten, Hilger and Fiebach2015) showed that structural and functional features of the brain do not overlap for accounting for intelligence differences.

Interestingly, after analyzing the relationships between functional connectomes and almost 300 behavioral/demographic measures in more than 450 individuals, Smith et al. (Reference Smith, Nichols, Vidaurre, Winkler, Behrens, Glasser and Miller2015) discovered one single mode of population covariation. These researchers made one straightforward parallelism between the identified population covariation and the general factor of intelligence (g) discovered more than a century ago by Charles Spearman. However, they underscore the fact that the covariation goes beyond intelligent behavior and includes further factors such as years of education, income, or life satisfaction. The general mode of positive (or negative) function results from the coordinated interactions among brain networks. This is definitely good news for those who assume that simple explanations are better than complex ones. In passing, note this key finding is highly consistent with TBC view.

Finally, the goal of ‘hunting’ specific genes responsible for the acknowledged genetic contribution to individual differences in intelligence has been elusive for more than twenty years. The recent study by Spain et al. (Reference Spain, Pedroso, Kadeva, Miller, Iacono, McGue and Simpson2015) is a key example. They compared 1409 individuals with IQs > 170 and 3253 non-overlapping controls, failing to find significant differences at the genetic level. This failure highlighted “the complex genetic architecture of intelligence”. However, somewhat surprisingly, using the UK Biobank (N = 112151) Davies et al. (Reference Davies, Marioni, Liewald, Hill, Hagenaars, Harris and Deary2016) have reported one successful attempt after applying genome-wide complex trait analysis (GCTA-GREML). They found genome-wide significant SNP-based associations in twenty genomic regions along with significant gene-based findings in forty-six regions. Nevertheless, as underscored in Danielle Posthuma presentation, novel approaches, based on uncovering the specific paths going from the genome to behavior, are strongly required.

Questions & Answers

I made two questions to each participant on this International Seminar. Of course, these are questions that interested me after listening to their talks and maybe they do not capture further issues that might be also relevant to the field. I will divide this section into short subsections devoted to each presentation. Afterwards, I will describe their answer to a common question.

M^a Ángeles Quiroga

Taking into account that she registered video game performance across several years, observing that volunteers required progressively less time for completing the same video game, my first question was: is there some sort of Flynn effect also affecting video game performance? The answer was straightforward (and predictable): not sure. More data are required for providing reliable responses.

The second question was based on the demonstrated fact that playing video games requires intelligence (in the classic sense), as shown in the thought-provoking report by Quiroga et al. (Reference Quiroga, Escorial, Román, Morillo, Jarabo, Privado and Colom2015): is it possible to enhance intelligence (the general factor, to be more specific) just playing a lot on a regular basis? The answer was ‘probably not’: maybe in the long run, playing video games of diverse kinds may promote the abilities and skills belonging to the intelligence realm, but the available evidence does not support this expectation.

Douglas K. Detterman

The main thesis of his presentation was this: teachers account for 1% –7% of the total variance in education, whereas the students’ intellectual level accounts for much of the 90% of variance associated with learning outcomes. Therefore, problems encountered for improving education might be explained by the (wrong) focus on schools and teachers instead on the (really) important variable (the student).

However, if this is the case, it is difficult to understand why different countries with the same average intelligence level show large performance differences on international assessments of school skills, such as PISA. Finland and Spain are good examples. Both countries show exactly the same average IQ level (97) (Lynn & Vanhanen, Reference Lynn and Vanhanen2012), but the first shows a PISA score of 523 while Spain shows a PISA score of 477 (the OCDE average score is 500).

The school system may have a great impact on developing countries, as demonstrated by the excellent study by Flores-Mendoza et al. (Reference Flores-Mendoza, Mansur-Alves, Ardila, Rosas, Guerrero-Leiva, Lucio-Gómez Maqueo and Burga León2015): the socioeconomic status (SES) of schools exerted a large impact on PISA scores. The effect of SES over school outcomes was higher for schools than for students. However, these sharp differences hardly apply to developed countries such as Finland and Spain. Detterman did not provide satisfactory answers to the core question because he thought there might be hidden relevant information.

My second question was straightforward: can we replace teachers by computerized courses and, therefore, save a big piece of public money? Now the answer was quick: yes, we can. No matter how you teach, because the relevant variable is the student. We must find novel ways for dealing with this reality.

Stuart Ritchie

His presentation focused on cognitive and physical decline in old age using a short-term longitudinal dataset from the Lothian Birth Cohorts. An interesting result was this: there is no a generalized decline. Some individuals do show cognitive but not physical decline (and vice versa). This is certainly shocking because it suggests some sort of dissociation between body and mind.

Nevertheless, my first question was: why the classic distinction between fluid and crystallized intelligence is neglected here? It is widely recognized that these two cognitive abilities show distinguishable trajectories across the life span (Hunt, Reference Hunt2011a). Talking about cognitive decline may hide different paths for these abilities, in the same sense as the mind-body dissociation noted above. This was Ritchie answer: only a short period of time was considered and, therefore, the evidence is clearly insufficient to detect those probable different paths for these cognitive abilities.

The second issue was also highly relevant: it is not possible to predict change based on the pretest measures (baseline). This was the rationale for my next question: are random effects responsible for this prediction failure? Jensen (Reference Jensen, Sternberg and Grigorenko1997) and Pinker (Reference Pinker2002) inspired this question. Both seem to converge on the suggestion that non-shared variance can be attributed to continuously changing random effects. Ritchie accepted this possibility: you cannot extract relevant signals from noise.

Danielle Posthuma

As said above, the hunt for genes accounting for the heritability estimates of intelligence has been disappointing. Posthuma acknowledged the fact, providing interesting examples from his own research group.

My first question was related with TBC view: relating the genome with behavior skipping the brain sounds like an inefficient research strategy, to say the least. Therefore, endophenotypes such as the brain must be incorporated into the big picture for a proper understanding of the genes-behavior relationships. Posthuma agreed with this view.

The second question was not a real question. Researchers acknowledge the fact that the effect of individual genes must be minuscule. This may help to explain the repeated failure underscored before. But perhaps there is also some sort of Butterfly effect: meager genetic differences from the outset (from a developmental perspective) can make a huge difference at latter developmental stages. This perspective fits the Dickens-Flynn model, which underscores that minor genetic advantages at the individual level might allow capturing relevant environmental factors for intelligence development in the long run (Flynn, Reference Flynn2007). This possibility must be investigated using a developmental perspective, as recently suggested by Johnson (Reference Johnson2013).

Emiliano Santarnecchi

His presentation was devoted to the neuroscience of intelligence, showing findings derived from functional data mainly obtained at rest. The brain shows spontaneous activity and there are large individual differences in how regions across the cortex are connected. Thus, for instance, high IQ individuals are more resilient to random and targeted attacks to their networks than low IQ individuals (Santarnecchi, Rossi, & Rossi, Reference Santarnecchi, Rossi and Rossi2015). He also presented evidence related with the problem of how can we enhance brain power using transcranial magnetic stimulation. There is still a lot of work remaining, but available results are encouraging.

My two questions were somewhat related. I firstly asked how the presented functional evidence could be integrated with the available structural results. Based on the meta-analysis described above (Basten et al., Reference Basten, Hilger and Fiebach2015) that revealed non-overlapping findings for structural and functional correlates of intelligence differences, I asked for probable ways of combining and making sense of both lines of evidence. However, the response was far from straightforward. Again, more research is required.

The second question derived from the observation that connectivity at rest shows greater reliability than task-evoked connectivity. This suggests that it may be more likely to find commonalities between the former connectivity and structural connections computed after extracting signals from diffusion data (Pineda-Pardo et al., Reference Pineda-Pardo, Martínez, Román and Colom2016). Santarnecchi agreed with this perspective, but he also recognized that the suggested multimodal approach would be highly complex. We already know this is the case, but we are also making progress in the right direction (Chamberland, Bernier, Fortin, Whittingstall, & Descoteaux, Reference Chamberland, Bernier, Fortin, Whittingstall and Descoteaux2015).

Adam Chuderski

Chuderski’s presentation was devoted to the problem of how individual differences in brain waves, especially their coupling, may account for observed intelligence differences (Chuderski & Andrelczyk, Reference Chuderski and Andrelczyk2015). The analysis of Theta/Gamma couplings provides findings supporting the view that larger numbers of gamma cycles within a given theta cycle is related with higher intelligence levels. My first question asked about the possibility of increasing (undesirable) noise within the system at higher levels of gamma activity. However, the answer rejected the issue relying on the available computer simulations. Nevertheless, it remains to be seen if these simulations properly capture the realities of the human brain.

The second question was related with what we know about complex systems functioning. We may be tempted to experimentally modify the way brain waves are coupled in, say, low IQ individuals showing less gamma cycles within a given theta cycle. Modifying gamma activity is possible using neurofeedback (Keizer, Verschoor, Verment, & Hommel, Reference Keizer, Verschoor, Verment and Hommel2010), and, therefore, the goal seems doable. However, as noted before, the brain is a complex system and, therefore, it is far from clear how modifying one single factor may stimulate a chain reaction with unclear consequences. No straightforward response was provided.

Norbert Jausovec

In this final presentation of the seminar, Jausovec described three studies devoted to the exciting possibility of increasing intelligence performance using cognitive and brain stimulation. My two questions had a methodological flavor. Firstly, I focused on the sex variable, asking about the relevance of controlling for the acknowledged average sex difference in brain volume when studying brain connectivity. There are reports showing sex differences in brain connectivity (Ingalhalikar et al., Reference Ingalhalikar, Smith, Parker, Satterthwaite, Elliott, Ruparel and Verma2014), but the observation vanishes when individual differences in brain volumes are controlled (Hanggi, Fovenyi, Liem, Meyer, & Jancke, Reference Hanggi, Fovenyi, Liem, Meyer and Jancke2014). Therefore, we must be very careful when studying men and women from a neuroscience perspective (Escorial et al., Reference Escorial, Román, Martínez, Burgaleta, Karama and Colom2015).

The second question derived from the discussion related with the source of the remarkable individual differences in working memory capacity that may support intelligence differences. As discussed by Bays (Reference Bays2015) available evidence fits with the perspective that working memory performance is influenced in a continuous way by increased memory load. Noise in neural activity degrades the stability/reliability of internal representations for further processing (Colom, Abad, Quiroga, Shih, & Flores-Mendoza, Reference Colom, Abad, Quiroga, Shih and Flores-Mendoza2008; Martínez et al., Reference Martínez, Burgaleta, Román, Escorial, Shih, Quiroga and Colom2011), which contrasts with the view that working memory capacity is limited by a fixed number of slots (Luck & Vogel, Reference Luck and Vogel2013). Jausovec was silent regarding this debate.

As advanced at the beginning of this section, the final question was common to all presenters:

All shared the idea of using the money for financing top scientists worldwide for finding answers to the next question:

Detterman led this common perspective. In some sense, the main goal is to replicate previous scientific efforts such as the Manhattan or the Apollo Projects. Having the best minds focused on solving a single, albeit complex, problem seems an efficient research strategy for a quick advance in our current knowledge related with the above key question.

I also adhere to this use of the money provided by Gates. Indeed, in the editorial note published two years ago (Colom, Reference Colom2014a) I made a metaphoric comparison with the Apollo Program: “we may wonder if we can make the voyage from the earth to the brain to increase our understanding of what it means to be more or less intelligent”. The first director of NASA’s Manned Spacecraft Center (Robert R. Gilruth) acknowledged the fabulous challenge involved in terms of people and technology, but he also was strongly prone to pursue the goal all the way. So I am.

In the farewell editorial note published by D. K. Detterman after being editor of the journal ‘Intelligence’ for four decades he wrote: “from very early, I was convinced that intelligence was the most important thing of all to understand, more important than the origin of the universe, more important than climate change, more important than curing cancer, more important than anything else. That is because human intelligence is our major adaptive function and only by optimizing it will we be able to save ourselves and other living things from ultimate destruction. It is as simple as that”.

We need shared efforts for enhancing our knowledge. The XXI Century will see increased numbers of scientists sharing their resources devoted to key research challenges. The reconciliation between humans and nature can only be achieved making a wise use of our main psychological faculty. This may be not an easy target, but I strongly think it is vital, literally.