Integrated Intelligence from Distributed Brain Activity

Abstract

Fluid intelligence tests predict success in many activities, suggesting cognitive mechanisms of broad importance.We propose a core process of attentional integration. Complex problems must be segmented into simpler parts. Attention to each part integrates cognitive fragments into a computational structure.Fluid intelligence is linked to the brain’s multiple-demand (MD) system, defined by common activity across different cognitive demands. Across the brain, MD patches shows anatomical and physiological properties adapted to attentional integration.Neurophysiology of putative MD regions shows adaptive coding of task-relevant information. Suiting attentional integration, many neurons show conjunctive coding (e.g., binding cognitive operations to their target objects).In broad outline, these results suggest how distributed brain activity builds organized cognition. How does organized cognition arise from distributed brain activity? Recent analyses of fluid intelligence suggest a core process of cognitive focus and integration, organizing the components of a cognitive operation into the required computational structure. A cortical ‘multiple-demand’ (MD) system is closely linked to fluid intelligence, and recent imaging data define nine specific MD patches distributed across frontal, parietal, and occipitotemporal cortex. Wide cortical distribution, relative functional specialization, and strong connectivity suggest a basis for cognitive integration, matching electrophysiological evidence for binding of cognitive operations to their contents. Though still only in broad outline, these data suggest how distributed brain activity can build complex, organized cognition. How does organized cognition arise from distributed brain activity? Recent analyses of fluid intelligence suggest a core process of cognitive focus and integration, organizing the components of a cognitive operation into the required computational structure. A cortical ‘multiple-demand’ (MD) system is closely linked to fluid intelligence, and recent imaging data define nine specific MD patches distributed across frontal, parietal, and occipitotemporal cortex. Wide cortical distribution, relative functional specialization, and strong connectivity suggest a basis for cognitive integration, matching electrophysiological evidence for binding of cognitive operations to their contents. Though still only in broad outline, these data suggest how distributed brain activity can build complex, organized cognition. Organized cognition of any kind arises from widely distributed brain activity. An immediate question is how such activity is integrated, allowing just the right cognitive contents to be combined in just the right way for current purposes. Though much is certainly unknown, an outline view of the relevant psychological and physiological mechanisms is beginning to appear. In this opinion article, we describe recent progress towards a whole-brain understanding of cognitive integration. We begin with recent work on the cognitive mechanisms of fluid intelligence (see Glossary). Theoretical accounts of fluid intelligence focus on processes of cognitive control [1.Duncan J. et al.Intelligence and the frontal lobe: the organization of goal-directed behavior.Cognit. Psychol. 1996; 30: 257-303Crossref PubMed Scopus (0) Google Scholar, 2.Kane M.J. Engle R.W. The role of prefrontal cortex in working memory capacity, executive attention, and general fluid intelligence: an individual-differences perspective.Psychon. Bull. Rev. 2002; 9: 637-671Crossref PubMed Google Scholar, 3.Unsworth N. Robison M.K. A locus coeruleus-norepinephrine account of individual differences in working memory capacity and attention control.Psychon. Bull. Rev. 2017; 24: 1282-1311Crossref PubMed Scopus (28) Google Scholar] and cognitive integration [4.Wilhelm O. et al.What is working memory capacity, and how can we measure it?.Front. Psychol. 2013; 4: 433Crossref PubMed Scopus (104) Google Scholar] and, based on recent findings, we suggest a synthesis of these two approaches. Results from brain imaging [5.Bishop S.J. et al.COMT val158met genotype affects recruitment of neural mechanisms supporting fluid intelligence.Cereb. Cortex. 2008; 18: 2132-2140Crossref PubMed Scopus (0) Google Scholar, 6.Duncan J. et al.A neural basis for general intelligence.Science. 2000; 289: 457-460Crossref PubMed Scopus (721) Google Scholar, 7.Prabhakaran V. et al.Neural substrates of fluid reasoning: an fMRI study of neocortical activation during performance of the Raven's Progressive Matrices Test.Cognit. Psychol. 1997; 33: 43-63Crossref PubMed Google Scholar] and lesion [8.Woolgar A. et al.Fluid intelligence loss linked to restricted regions of damage within frontal and parietal cortex.Proc. Natl. Acad. Sci. U. S. A. 2010; 107: 14899-14902Crossref PubMed Scopus (118) Google Scholar] studies relate fluid intelligence to a well-known control network in the brain, which previously we have called the multiple-demand (MD) system [9.Duncan J. The multiple-demand (MD) system of the primate brain: mental programs for intelligent behaviour.Trends Cogn. Sci. 2010; 14: 172-179Abstract Full Text Full Text PDF PubMed Scopus (853) Google Scholar,10.Duncan J. The structure of cognition: attentional episodes in mind and brain.Neuron. 2013; 80: 35-50Abstract Full Text Full Text PDF PubMed Scopus (239) Google Scholar]. We describe recent studies on the detailed anatomy and physiology of MD activity and how they begin to illuminate the physiological underpinning of cognitive control and integration. In broad outline, these findings suggest how distributed brain activity builds organized cognition. We conclude with some of the many questions that this scheme raises for future work. A fundamental psychometric discovery is positive manifold: to some extent, all tests of different cognitive abilities tend to have positive correlations [11.Spearman C. General intelligence, objectively determined and measured.Am. J. Psychol. 1904; 15: 201-293Crossref Google Scholar,12.Spearman C. The Abilities of Man. Macmillan, 1927Google Scholar], even those that on the surface are dissimilar. In his foundational work, Spearman [11.Spearman C. General intelligence, objectively determined and measured.Am. J. Psychol. 1904; 15: 201-293Crossref Google Scholar] proposed that some general or g factor contributes to success in any task. If this model is to tests to be of the in a be these tests that have the correlations with a of are problems et for Raven's Progressive Matrices and for and with the The for and Scholar], for the Abilities Scholar], The in such tests called we to positive the broad of fluid intelligence tests to predict success in many of from tests to cognitive mechanisms of in mental We begin with of fluid intelligence based on cognitive or executive work fluid intelligence to frontal [1.Duncan J. et al.Intelligence and the frontal lobe: the organization of goal-directed behavior.Cognit. Psychol. 1996; 30: 257-303Crossref PubMed Scopus (0) Google J. intelligence and the frontal The J. et al.Fluid intelligence frontal 33: PubMed Scopus (0) Google Scholar], of control have in fluid intelligence in working memory in the of M.J. Engle R.W. The role of prefrontal cortex in working memory capacity, executive attention, and general fluid intelligence: an individual-differences perspective.Psychon. Bull. Rev. 2002; 9: 637-671Crossref PubMed Google Scholar], and N. Robison M.K. A locus coeruleus-norepinephrine account of individual differences in working memory capacity and attention control.Psychon. Bull. Rev. 2017; 24: 1282-1311Crossref PubMed Scopus (28) Google M.J. et the mind and an study of working memory and executive control in Sci. 18: PubMed Scopus Google Scholar], mental M.J. Engle R.W. The role of prefrontal cortex in working memory capacity, executive attention, and general fluid intelligence: an individual-differences perspective.Psychon. Bull. Rev. 2002; 9: 637-671Crossref PubMed Google Scholar], and Though this broad focus is and the to frontal is of control such in working memory Scholar], attentional and a to and in Rev. PubMed Google Scholar, J. mechanisms of Rev. 18: PubMed Google Scholar, Attention to and of of for Scholar], et al.Neural the of PubMed Scopus Google A. et and of executive and their to a Psychol. 2000; PubMed Scopus (0) Google Scholar], and the on of cognition. their they of be to even a such and a to the A different on cognitive control from work in intelligence, to the programs of et A. et of a of Rev. Scopus Google Scholar]. this work problems be in a they must be into simpler with these simpler in to success on the A organized a is in is to the of to and to the question of how to the of the to to beginning with the of to the can of a and the and a to a in a of Scopus (0) Google Scholar]. that a core of must be cognitive based on of the and to the a must be organized into parts. work et al.What intelligence a account of the in the Progressive Matrices Rev. PubMed Google Scholar] the of cognitive in fluid intelligence In a recent study J. et and in fluid Natl. Acad. Sci. U. S. A. 2017; PubMed Scopus Google Scholar], we problems that this in in each of the this or or or the working memory of problems the their into a allowing each part of the to be The attention is on a part of the (e.g., just on the with on a fluid intelligence to these problems that the only we problems that with each part of the in Though problems the those that be by attention to a part in the all accounts of fluid intelligence based on cognitive control have with an account based on cognitive integration [4.Wilhelm O. et al.What is working memory capacity, and how can we measure it?.Front. Psychol. 2013; 4: 433Crossref PubMed Scopus (104) Google A. a binding of is for fluid Scopus (0) Google Scholar]. to integration the process in fluid intelligence is binding the different components of a cognitive process or In this is to is to to part of the in (e.g., to that the is on in each of the must be to within each and in the a The of and right must be to an of the to be and for the to is by to each attention, cognitive fragments must be into the and these computational are the a is segmented into each a to the The for attentional integration the of each of components into just the right computational structure is in fluid intelligence We that attentional integration the of all organized cognition to fluid intelligence tests predict success in such a of into is in even just the right cognitive organized in just the right must be for we suggest that accounts of fluid intelligence based on control and integration two of the attention, to and integration are all of brain activity that and the contents of an cognitive accounts of fluid intelligence within a view of how in general be the to be in a cognitive structure be in widely distributed brain activity. attention to a integration of activity across cortical and brain that different properties and J. et brain activity in 1997; PubMed Scopus Google et a model for attention in Psychol. PubMed Google Scholar]. a cognitive structure such a relevant activity be distributed across much or of the for an integration with to neural activity is in current cognition. with regions of brain activity linked to specific cognitive brain imaging studies show activity in a widely distributed MD a of brain regions in which activity with any kind of cognitive J. The multiple-demand (MD) system of the primate brain: mental programs for intelligent behaviour.Trends Cogn. Sci. 2010; 14: 172-179Abstract Full Text Full Text PDF PubMed Scopus (853) Google Scholar,10.Duncan J. The structure of cognition: attentional episodes in mind and brain.Neuron. 2013; 80: 35-50Abstract Full Text Full Text PDF PubMed Scopus (239) Google J. regions of the frontal by cognitive 2000; Full Text Full Text PDF PubMed Scopus Google et in regions of frontal and parietal cortex.Proc. Natl. Acad. Sci. U. S. A. 2013; PubMed Scopus Google for an and for in the 2013; PubMed Scopus Google et of the multiple-demand network during PubMed Scopus Google MD activity is in regions on the frontal in and the in frontal and within and to of the and in a the occipitotemporal the of brain MD activity linked to cognitive control J. An adaptive coding model of neural in prefrontal Rev. PubMed Scopus Google The cognitive control cortical regions with PubMed Scopus Google Scholar] and integration The cognitive control cortical regions with PubMed Scopus Google activity in attention, and on Cogn. 14: PubMed Scopus (239) Google Scholar]. A core role in fluid intelligence [9.Duncan J. The multiple-demand (MD) system of the primate brain: mental programs for intelligent behaviour.Trends Cogn. Sci. 2010; 14: 172-179Abstract Full Text Full Text PDF PubMed Scopus (853) Google a account of the general factor of Scopus (0) Google M.J. and of for the of cognitive Rev. PubMed Scopus Google Scholar] is by strong MD activity during work on fluid intelligence problems [5.Bishop S.J. et al.COMT val158met genotype affects recruitment of neural mechanisms supporting fluid intelligence.Cereb. Cortex. 2008; 18: 2132-2140Crossref PubMed Scopus (0) Google Scholar, 6.Duncan J. et al.A neural basis for general intelligence.Science. 2000; 289: 457-460Crossref PubMed Scopus (721) Google Scholar, 7.Prabhakaran V. et al.Neural substrates of fluid reasoning: an fMRI study of neocortical activation during performance of the Raven's Progressive Matrices Test.Cognit. Psychol. 1997; 33: 43-63Crossref PubMed Google Scholar] and of fluid intelligence with MD damage [8.Woolgar A. et al.Fluid intelligence loss linked to restricted regions of damage within frontal and parietal cortex.Proc. Natl. Acad. Sci. U. S. A. 2010; 107: 14899-14902Crossref PubMed Scopus (118) Google A. et al.Fluid intelligence is by the multiple-demand system the PubMed Scopus Google Scholar, et for general intelligence and executive by lesion PubMed Scopus Google Scholar, J. et neural system for general intelligence by lesion Natl. Acad. Sci. U. S. A. 2010; 107: PubMed Scopus Google Scholar]. In this we findings on MD anatomy and physiology to the computational of attentional integration. Though activity for many anatomy this we to the and data of the et al.A cognitive core defined in Cortex. 30: PubMed Scopus (0) Google Scholar]. In the brain much of cortical data are to the cortex of each into regions et al.A of PubMed Google Scholar]. In a of we the of cognitive working memory and to MD regions and their The results the anatomy of MD suggesting Across the cortex we nine defined patches of MD distributed across frontal, parietal, and cortex The MD patches into individual which we into a core with activity and a MD patches on the frontal the just to the frontal the and regions of the frontal and a of from the frontal into the frontal cortex An frontal on the frontal surface In the parietal the MD is in the of the with an in parietal cortex and the that of we MD activity in a the of and cortex this of MD we can functional of MD have many of functional (e.g., et al.A core system for the of Full Text Full Text PDF PubMed Scopus Google Scholar, et and in Cortex. PubMed Scopus Google Scholar, A. et Full Text Full Text PDF PubMed Scopus Google with Across the of MD shows of activity for The results of the is a strong of with activity for and for across all MD regions the is with activity and with this of even differences are Though the MD system is a the of this recruitment from to (see S. et recruitment of the multiple-demand system with and Cogn. PubMed Scopus Google data to of functional studies define a control with the MD system that is defined by activation data et organization of the cortex by functional PubMed Scopus Google Scholar]. on shows connectivity for all of and by of core to with a previously by et et the functional network PubMed Scopus (0) Google Scholar] from the that all regions in MD core within the and, within that are by are across Recent analyses of brain connectivity have structure O. Complex brain of and functional Rev. PubMed Scopus Google Scholar], with strong connectivity within and and suggest that MD regions are the brain’s widely et the brain’s 2010; PubMed Scopus Google et for in functional brain 2013; Full Text Full Text PDF PubMed Scopus Google Scholar], with to brain et connectivity for adaptive 2013; PubMed Scopus Google et and of frontal in Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar]. accounts fluid intelligence to this MD connectivity et connectivity of prefrontal cortex cognitive control and PubMed Scopus Google Scholar], a recent of Cogn. Sci. Full Text Full Text PDF PubMed Scopus Google Scholar]. In with such results suggest that the MD system is for attentional integration in the brain, a system to many widely distributed parts. we different of are into different MD for functional MD to in many cognitive that all of can be and in the and required by a cognitive connectivity MD in regions of the a for and integration and an for the strong of across many different demands. A of the findings is Though analyses on MD activity in of the and and in specific regions of the Though are unknown, is that attentional integration is cortical and activity. If MD regions are to cognitive the contents of these episodes must be in the of MD In brain MD of different of a current of relevant and S. et coding for in the PubMed Scopus Google Scholar, et in the Full Text Full Text PDF PubMed Scopus Google Scholar, A. et coding of task-relevant in PubMed Scopus (0) Google Scholar]. In to such or MD of relevant contents is A. et is to from prefrontal cortex activity Cogn. 30: PubMed Scopus Google Scholar] A. et of and in the brain: of Cogn. PubMed Scopus Google Scholar]. a role in attention, MD is relevant to a current A. et of in frontal, parietal and PubMed Scopus (0) Google Scholar, J. et attention in to task-relevant Cogn. 2017; PubMed Scopus Google Scholar, J. of based on in regions of PubMed Scopus Google Scholar]. findings, MD is A. et coding of task-relevant in PubMed Scopus (0) Google Scholar], to the of findings from in the and are imaging data suggest a MD network that of the brain et fMRI activation for in PubMed Scopus (0) Google et al.A putative multiple-demand system in the PubMed Scopus Google Scholar], regions of frontal, frontal, and parietal cortex. In neurons in these regions show properties of findings, with of many neurons the of a current and their properties to relevant et of in the primate prefrontal PubMed Scopus Google Scholar, of in parietal PubMed Scopus Google Scholar, S. et of neural by attention in the prefrontal 2002; PubMed Scopus Google Scholar, of of by primate prefrontal to relevant PubMed Google Scholar, et of attention and across prefrontal PubMed Scopus Google Scholar, et control of cortex in a PubMed Scopus (28) Google Scholar]. these data suggest a distributed network with to task-relevant and network control of attention in the prefrontal and parietal PubMed Scopus Google Scholar, et during working PubMed Scopus Google Scholar, et correlations of and prefrontal cortex during attention PubMed Scopus Google Scholar, et and prefrontal cortex attention Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar]. for current the properties of MD neurons suggest a role in integration. we have of a cognitive relevant contents into the required and The of a how and be the required of how are linked to in a that of must be to such the an to the can be contents to is the computational of binding binding and the of in Scopus Google Scholar] and, for neural a for conjunctive coding of contents and In recent conjunctive coding in frontal and regions to the of neural activity by of et of by the of the of neural 2010; 4: PubMed Scopus Google et of in cognitive 2013; PubMed Scopus Google Scholar]. The many of If an must a of for many frontal neurons a can be of in memory in the prefrontal 2010; 30: PubMed Scopus Google et of primate prefrontal cortex and to Natl. Acad. Sci. U. S. A. 2017; PubMed Scopus (0) Google Scholar]. If a that the for a target activity in the and the this can be of activity the target et coding for cognitive control in prefrontal 2013; Full Text Full Text PDF PubMed Scopus Google Scholar]. In a study of frontal activity during et in the prefrontal cortex of future in Full Text Full Text PDF PubMed Scopus Google Scholar], an of a from a to a the individual frontal be for the of just the just the or just the in the In with the of cognitive integration, MD regions is to MD regions conjunctive binding the components of a cognitive operation into the computational structure. An from recent work et of episodes in frontal and parietal Cortex. 30: PubMed Scopus (0) Google Scholar] binding of cognitive operations to their target In this of on a and, on each the a by each or two the to these by and and on We data from two the during which the and a that they their and the the a the target or in two frontal and parietal cortex in parietal cortex each we two of to activity and we of activity across the or activity for different and correlations to each in in of in the that this to and frontal and parietal neurons that the of the a during or of the MD regions the cognitive operations of allowing the cognitive operations to be each N. et coding for in the prefrontal cortex.Proc. Natl. Acad. Sci. U. S. A. 2008; PubMed Scopus Google Scholar]. for of and to these different we positive the of positive correlations different cognitive and that some general or g factor contributes to success in any cognitive activity [11.Spearman C. General intelligence, objectively determined and measured.Am. J. Psychol. 1904; 15: 201-293Crossref Google Scholar,12.Spearman C. The Abilities of Man. Macmillan, 1927Google Scholar]. is that g the attentional integration of the MD system and, in with we suggest that MD to cognition. the that the for positive is with a account of the general factor of Scopus (0) Google of Cogn. Sci. Full Text Full Text PDF PubMed Scopus Google Scholar]. any the control structure is a of MD MD activity with activity in in operations of this and that the of the computational structure on all and their In many fluid intelligence to with different of brain distributed across many brain regions et of intelligence Rev. 2010; PubMed Scopus Google U. et are a of functional and brain imaging studies on Scopus Google Scholar]. In a recent for intelligence with the and of in from the of et and neurons with PubMed Scopus Google Scholar]. many neural properties be across regions of the and in the be of in MD regions and In this a distributed of neural the general process of cognitive their with whole-brain functional such of functional et of functional brain with PubMed Scopus Google intelligence is with brain network PubMed Google Scholar]. is much the core cognitive mechanisms required in fluid intelligence the of the MD system, and of individual differences to positive these are to be The role of on positive and a that is fluid and fluid intelligence current intelligence the of a and The of and brain and the of Scholar]. The two are is during and a of to a of with of this to many to success in and to positive A is by the of in that in or have positive on in et al.A model of general intelligence: the positive of intelligence by Rev. PubMed Scopus Google et and cognitive during and Sci. 2017; PubMed Scopus Google Scholar]. this must be to for must mental and in many future If positive the broad to cognitive is by many of are by the integration we the of cognitive activity and and the of attentional capacity by of positive a core question is cognition and in intelligence (e.g., A. of a is into on the basis of of the structure of the and within is that how to can be into how a can be in a or we in a to a In the brain, that current cognition is distributed across brain for be based a proposed in the et neural and computational of Rev. 2017; 18: PubMed Google Scholar], and are linked to components of the network et of the brain’s 2010; Full Text Full Text PDF PubMed Scopus Google Scholar]. MD activity in to cognitive we to how of into this this is of the connectivity of MD regions and that have in the MD In intelligence, in an working of a structure of and cognition this is for such to or a only a can be in neural any with the of the structure in for the the current focus of attention must to the for a to progress to a by can a for an The is of recent accounts of working neural with working memory in prefrontal a coding Cogn. Sci. Full Text Full Text PDF PubMed Google et and working Full Text Full Text PDF PubMed Google Scholar]. is how the focus of attention in cognition can be within a complex, of the A the well-known capacity of in and Google Attention and Scholar]. on MD activity an basis for such and, have linked capacity to the of frontal and parietal cortex J. intelligence and the frontal The Attention to and of of for S. et al.A network model and physiological data during Natl. Acad. Sci. U. S. A. PubMed Scopus Google J. of in the Cogn. Sci. 9: Full Text Full Text PDF PubMed Scopus Google Scholar]. in in frontal and parietal is of different et of a attentional in a prefrontal 2013; 80: Full Text Full Text PDF PubMed Scopus Google Scholar], working memory et al.Neural substrates of cognitive capacity Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar], or components et of episodes in frontal and parietal Cortex. 30: PubMed Scopus (0) Google S. mechanisms of and cognitive capacity in the prefrontal PubMed Scopus Google Scholar]. work is to the physiological basis of this In the system, capacity in are to arise a process of or et of neurons in cortex during 80: PubMed Google Scholar, et mechanisms attention in and PubMed Google Scholar, The model of Full Text Full Text PDF PubMed Scopus Google Scholar]. In such each to the activity of a to a to the properties of this with in the activity to a the of for any can be in the of prefrontal neurons control processes working PubMed Scopus Google A. of attention on relative target during PubMed Scopus Google Scholar], the that is a general in MD cortex. neural have of working memory and cognitive control (e.g., et by how the network 2017; Full Text Full Text PDF PubMed Scopus Google and in a recent is the basis for working memory capacity A model of working Full Text Full Text PDF PubMed Scopus Google Scholar]. work is to be to the attentional of MD activity and cognitive account of cognitive integration much an of the an outline of how distributed brain activity can cognition. we the we to detailed physiological study (see different MD regions Across the MD system, we of the of and during The different of MD regions functional to cognitive integration, in the is of the of the to and how task-relevant arises and is distributed across the network for electrophysiological in MD in or for does MD activity across brain this for electrophysiological questions that different of a cognitive and the role of (e.g., across brain are of cognition combined into we of how brain activity in of a of for a question is immediate cognitive activity and of and their is the role of MD in and is of and cognitive operations are imaging the for studies of How different MD regions Across the MD system, we of the of and during The different of MD regions functional to cognitive integration, in the is of the of the to and how task-relevant arises and is distributed across the network for electrophysiological in MD in or for How does MD activity across brain this for electrophysiological questions that different of a cognitive and the role of (e.g., across brain How are of cognition combined into we of how brain activity in of a of for a question is immediate cognitive activity and of and their is the role of MD in and is of and cognitive operations are imaging the for studies of work by the the a to and a to the in psychometric tests of Though many tests and fluid intelligence tests can or a distributed of cortical regions of activation with many different cognitive demands. a neural in which is by a of (e.g., a only a in a memory

Topics & Concepts