3.10 The Parieto-frontal Integration Theory (PFIT)
In December 2003, I hosted an invited symposium at the annual meeting of the International Society for Intelligence Research (ISIR). It was the first time imaging researchers came together to discuss intelligence studies. In addition to myself, participants included Jeremy Gray, Vivek Prabhakaran, Rex Jung, Aljoscha Neubauer, and Paul Thompson. With the exception of Aljoscha Neubauer, it was the first time I met these researchers in person. Rex Jung’s presentation was a compelling review of several studies that emphasized the distributed nature of brain areas associated with intelligence. Based on his clinical background as a neuropsychologist and his MRS research on white matter and IQ, he also emphasized the importance of white matter connections among the salient brain areas. It was apparent that he and I had similar interests so we undertook a comprehensive review of the entire brain imaging/intelligence literature. It took us over two years to write the review, which was published in 2007 along with commentaries from other researchers (Haier & Jung, 2007; Jung & Haier, 2007).
From our first PET study of only eight subjects in 1988 to the larger fMRI studies through 2006, there were 37 imaging studies of intelligence from different research groups around the world. Given the wide disparity of methods and measures, and the number of potential brain areas involved, a typical meta-analysis was not appropriate. Instead, we followed a method used to review the emerging literature from cognitive neuroimaging studies (Cabeza & Nyberg, 2000). We reviewed structural MRI results, PET results, and fMRI results. We focused on findings common among studies irrespective of different imaging and assessment methods. Several brain areas were common in 50% or more of the 37 studies. This may seem a rather weak proportion, but it is similar to the proportions found in the Cabeza and Nyberg review of well-controlled cognitive experiments.
The salient brain areas we identified were distributed throughout the brain, but mostly were in parietal and frontal areas. We called our model the Parieto-frontal Integration Theory (PFIT) of Intelligence. Note that “Integration” emphasizes that communication among the salient areas was key to the model because we have always recognized that identifying specific brain areas was only the beginning of a useful brain model of intelligence. Understanding the temporal and sequential interactions among networks that link the areas would be key. The illustration in Figure 3.7 shows all the areas we included in the model. Animation 3.1 on this book’s website (www.cambridge.org/us/academic/subjects/psychology/ cognition/neuroscience-intelligence) shows the PFIT areas in 3D.
Figure 3.7 The Parieto-frontal Integration Theory (PFIT) showing brain areas associated with intelligence (courtesy Rex Jung). (See also Animation 3.1 on the website, www.cambridge.org/us/academic/subjects/psychology/cognition/neuroscience-intelligence.)
The circles in Figure 3.7 show brain areas and the numbers refer to the standard Brodmann area (BA) nomenclature (Brodmann, 1909). We proposed that these areas define a general brain network and subnetworks that underlie intelligence. Most of the areas are in frontal and parietal lobes, some in the left hemisphere (blue circles) and some in both hemispheres (red circles). A major white matter tract of fibers (yellow arrow) connects the frontal and parietal lobes like a super highway. It’s called the arcuate fasciculus and we have proposed that it is an important tract for intelligence.
The brain areas in our model represent four stages of information flow and processing while engaged in problem-solving and reasoning. In stage 1, information enters the back portions of the brain through sensory perception channels. In stage 2, the information then flows forward to association areas of the brain that integrate relevant memory, and in stage 3 all this continues forward to the frontal lobes that consider the integrated information, weigh options, and decide on any action, so in stage 4 motor or speech areas for action are engaged if required. This is unlikely to be a strictly sequential, one-way flow. Complex problems are likely to require multiple, parallel sequences back and forth among networks as the problem is worked in real time.
The basic idea is that the intelligent brain integrates sensory information in posterior areas, and then the information is further integrated to higher-level processing as it flows to anterior areas. The PFIT also suggests that any one person need not have all these areas engaged to be intelligent. Several combinations may produce the same level of general intelligence, but with different strengths and weaknesses for other cognitive factors. For example, two people might have the same IQ, or g level, but one excels in verbal reasoning, and the other in mathematical reasoning. They may both have some PFIT areas in common, but it is likely they will differ in other areas.
Cognitive studies show that some PFIT areas of the brain are related to memory, attention, and language, suggesting that intelligence is built on integrating these fundamental cognitive processes. Our hypothesis is that individual differences in intelligence, whether the g-factor or other specific factors, are rooted both in the structural characteristics of the specific PFIT areas and in the way information flows around these areas. Some people will have more gray matter in important areas or more white matter fibers connecting areas and some people will have more efficient information flow around the PFIT areas. These brain features lead some individuals to score higher on intelligence and mental ability tests, and other individuals to be less efficient, and less good at problem-solving. How the salient brain features may develop is a separate issue for future longitudinal studies of children and adolescents. In the next chapter we will see newer imaging methods that show millisecond changes in information flow throughout the brain so hypotheses about efficient information flow and intelligence can be tested.
While we were formulating the PFIT, we were unaware of a similar review published in a book chapter by two cognitive psychologists (Newman & Just, 2005). These authors also favored a distributed network for intelligence rather than a model concentrated only in the frontal lobe. Additionally, they noted the importance of white matter connections among brain areas. Efficient information flow and the importance of computational load were prominent features of their model. Independently, we arrived at similar views although we came from different perspectives. Their work is listed at the end of this chapter under Further Reading. I highly recommend it.
There is one more thing to mention. Many of the gray and white matter areas related to IQ first reported in these phase one studies that contributed to the PFIT appear to be under genetic control to some degree (Pol et al., 2002; Posthuma et al., 2002, 2003a; Thompson et al., 2001; Toga & Thompson, 2005) and we will discuss these and newer, even more compelling findings in the next chapter in studies that combine advanced genetic analysis, including DNA, and neuroimaging in very large samples.
Since our review of 37 studies was published in 2007, there have now been more than 100 additional imaging studies of intelligence from research groups all over the world, as more researchers appreciate the connections between general intelligence and fundamental cognitive processes. We refer to these post-2006 studies as phase two in the application of neuroimaging to intelligence research (Haier, 2009a). This new wave of studies includes many that are far more sophisticated with respect to large, representative samples, multiple measures of intelligence to estimate the g-factor, and advanced image-analysis techniques that include better anatomical measurement and localization methods. We will detail important phase two studies in the next chapter.
3.11 Einstein’s Brain
Before closing this chapter, let me draw your attention briefly to Einstein’s brain. It was removed after his death and preserved in a jar by a physician who kept it at home and then in his car as he moved across the country. He was reluctant to share it, but eventually samples were made available to researchers. The main findings (Witelson et al., 1999; Witelson & Harvey, 1999), not without technical issues that could influence interpretation of results (Galaburda, 1999; Hines, 1998), were that Einstein’s brain showed more tissue and more neuron-support cells in a post
erior part of the brain. This area was pretty much the same parietal area where the men showed correlations with IQ and the women didn’t (Haier et al., 2005). A detailed analysis of photographs of Einstein’s brain also suggested differences in frontal and parietal areas (Falk et al., 2013). Anything about how Einstein’s brain may differ from other brains is inherently interesting, but perhaps the most remarkable thing about his brain is that it is not all that remarkable from a purely anatomical analysis. In fact, at autopsy it is often the case that a person who had an IQ under 70 may have no remarkable anatomical brain features to distinguish it from brains of people with high IQs. This is why functional neuroimaging and quantitative image analysis have provided many new insights.
During the first phase of applying new medical neuroimaging technologies, intelligence researchers had limited access to expensive equipment and the first studies were characterized by small samples, single measures of intelligence, and rudimentary image-analysis methods that often ignored individual differences. Nonetheless, slow but steady progress from 1988 to 2006 allowed a literature review based on 37 studies that concluded there were a finite number of identifiable areas distributed across the brain where structure and/or function were related to scores on intelligence and reasoning tests. Phase two of imaging/intelligence studies builds on these findings with advanced methods and the latest progress is the focus of Chapter 4.
Chapter 3 Summary
This chapter laid out the early history of neuroimaging studies of intelligence, a period from 1988 to 2006 we refer to as phase one that indicated surprising findings.
From the first studies, it was apparent to most researchers that intelligence was not centered solely in frontal lobes, but instead involved networks distributed across the brain.
A surprising early finding was an inverse correlation between intelligence test scores and brain activity determined by glucose metabolic rate, suggesting a hypothesis that efficient information flow was an element of higher intelligence.
Imaging studies showed that not all brains worked the same way. Individual differences required examination rather than being ignored when group data were averaged.
Despite the limitations of phase one studies, some consistent results across studies suggested the Parieto-frontal Integration Theory of intelligence that emphasized both the structural and functional characteristics of specific brain areas and the connections among them.
Review Questions
1. Explain the difference between structural and functional neuroimaging.
2. What are the main differences between the PET and MRI technologies?
3. What is the basis for the brain efficiency hypothesis of intelligence?
4. What is the evidence about whether there is an “intelligence center” in the brain?
5. List key limitations of the early brain-imaging studies of intelligence.
Further Reading
Looking Down On Human Intelligence (Deary, 2000). This is a sophisticated and comprehensive account of intelligence research. Clearly written with wit and without jargon, it ranges from early thinkers and philosophers to the end of the twentieth century, including the early neuroimaging studies.
“The Parieto-Frontal Integration Theory (P-FIT) of intelligence: Converging neuroimaging evidence” (Jung & Haier, 2007). This is the original, somewhat technical review of 37 imaging/intelligence studies. It includes a broad range of commentaries from other researchers in the field (Haier & Jung, 2007).
“Human intelligence and brain networks” (Colom et al., 2010). This is a more general description of the PFIT model.
IQ and Human Intelligence (Mackintosh, 2011). This is a thorough textbook that covers all aspects of intelligence written by an experimental psychologist. It has a chapter that is a good summary of early imaging studies of intelligence (chapter 6).
“The neural bases of intelligence: A perspective based on functional neuroimaging” (Newman & Just, 2005). This chapter is clearly written and presents a brain model of intelligence similar to, but developed independently of, the PFIT.
Chapter Four
50 Shades of Gray Matter: A Brain Image of Intelligence is Worth a Thousand Words
There are more than enough brain-injured people in the modern world to permit resolution of every fundamental question concerning the human mind, could this material but be brought under adequate study.
(Ward C. Halstead, 1947, p. v)
The data are intriguing. The field is maturing. The pace is quickening. As intelligence research engages 21st century neuroscience, new hypotheses and new controversies are inevitable. What a terrific time to work in this field.
(Richard Haier, 2009a, p. 121)
Learning Objectives
How has neuroimaging revealed brain networks related to intelligence?
What is the empirical support for the PFIT framework?
Does the weight of evidence support a relationship between brain efficiency and intelligence?
Why is it difficult to predict intelligence test scores from brain images?
Does imaging research on intelligence differ from imaging research on reasoning?
Which brain structures share genes with intelligence test scores?
How have neuroimaging studies advanced the search for specific genes and brain mechanisms related to intelligence?
Introduction
Any lingering doubts that intelligence is a rich topic for neuroscience should be melted away given the genetic and neuroimaging studies described in the previous two chapters. If not, please suspend any remaining disbelief until the end of this chapter wherein even more compelling findings are described. They come from the most recent neuroimaging studies, including those obtained in conjunction with genetic methods. We include studies of adults and children, patients with brain damage, and introduce more advanced methods of brain image acquisition and analysis. These studies continue to intrigue and motivate researchers worldwide to push assessment technologies to even greater precision, increase the sample sizes to levels previously unimaginable, and offer new, testable hypotheses about intelligence and the brain. One word of caution: most of the studies in the last chapter and many in this chapter have sample sizes that are too small for conclusive interpretations. Remember our second law – no one study is definitive. As this field continues to mature, sample sizes are increasing rather dramatically. Over time, the weight of evidence always favors studies with sufficient sample sizes that maximize the stability of findings and minimize unreliable ones. This is especially so for the early studies seeking to identify candidate genes related to intelligence. I am including such studies for historical context and for illustrating how the weight of evidence evolves.
The PFIT framework described in Chapter 3 proposed that intelligence was related to 14 specific areas distributed throughout the brain (Jung & Haier, 2007; Haier & Jung, 2007). These areas formed a broad network of frontal–parietal communication along with subnetworks involving several other temporal and occipital areas. How information flowed through these networks was proposed as a basis for individual differences in mental abilities, and especially for the g-factor. The model also proposed that individuals with the same IQ might achieve their level of g from different combinations of PFIT areas. In other words, there may be multiple, even redundant, neuro-pathways to the g-factor just like there are multiple routes driving from New York City to Los Angeles. Efficient information flow, through whichever subnetworks are relevant for an individual, was hypothesized to relate to high g, and subnetworks of the PFIT were hypothesized to relate to a person’s pattern of mental ability strengths and weaknesses.
At the time the PFIT was proposed in 2007, testing these hypotheses was difficult. Methods of neuroimaging and analysis were limited with respect to their ability to assess structural or functional brain network connections and how well information was processed in networks during problem-solving. This state of affairs improved quickly and dramatically for intelligence research wi
th the application of new mathematical/statistical ways to assess connectivity among brain areas, new image-analysis techniques to assess the integrity of white matter transmission of information, and the use of the magneto-encephalogram (MEG) technology to assess regions of neuron activity dynamically every millisecond during the performance of a cognitive task. Adding to the increased pace of intelligence research, neuroimaging is now combined with genetic methods in several large-scale consortia. These advances are the focus of this chapter. There are at least 50 recent neuroimaging studies to choose from that illustrate the momentum of these advances in intelligence research. We cannot summarize them all, but let’s start with some key studies of brain network connectivity and what they found. These studies are presented in mostly chronological order so that the story is told as it has unfolded. All the studies reported in this chapter implicate many brain areas. I decided to include the major ones for completeness. You will get a sufficient feel for the general findings without memorizing these areas. It may be helpful to refer to the brain area maps in Figures 3.6 and 3.7 from the last chapter as you read this chapter.
4.1 Brain Networks and Intelligence
Every brain image is constructed from many thousands of small voxels. As explained in the last chapter, each voxel is assigned a value based on the type of imaging used. In the FDG PET studies described in the last chapter, the value was glucose metabolic rate. In structural MRI, the value can be density of gray or white matter. In fMRI, the value is based on blood flow. To determine how one brain area may be related to all other brain areas, correlations can be computed between any individual voxel, or group of voxels, defining a region of interest (ROI) and all other voxels (or ROIs) throughout the entire brain. The starting voxel is called the “seed.” Multiple seeds can be placed wherever the researchers wish depending on the hypothesis to be tested. The pattern of correlations indicates how the seed areas are connected to other brain areas. The connections are statistical and may or may not reflect actual anatomical connections.
The Neuroscience of Intelligence Page 14