The Neuroscience of Intelligence
Page 24
6.2 Cognitive Neuroscience of Memory and Super-Memory
In Chapter 1, we noted that one definition of intelligence is individual differences in the cognitive processes of learning, memory, and attention. Most cognitive neuroscience research does not include any assessment of intelligence as either an independent or dependent variable. As we reviewed in Chapter 4, results from any study of learning, memory, language, or attention may differ if participants are selected on the independent variable of high or low IQ or g-factor scores. As noted in Chapter 5, when intelligence is included as a dependent variable, as in the n-back training studies, the assessment typically is based on a single test score rather than on a latent variable extracted from a test battery. All this is the old bad news. The more recent good news is that cognitive psychologists are becoming more interested in the relationships among language, memory, attention, and intelligence. One area with considerable research is the relationship between working memory and the g-factor. In some psychometric studies they are empirically virtually identical (Colom et al., 2004; Kane & Engle, 2002; Kyllonen & Christal, 1990). In other studies they are overlapping but separate constructs (Ackerman et al., 2005; Conway et al., 2003; Kane et al., 2005). Imaging research suggests some overlap in brain areas for both (Colom et al., 2007) and both may have genes in common (Luciano et al., 2001; Posthuma et al., 2003a), but these issues are not yet settled (Burgaleta & Colom, 2008; Colom et al., 2008; Thomas et al., 2015). The ultimate goal is to understand how intelligence may integrate fundamental cognitive processes like memory and attention and the way they influence language and learning. This will require cooperation among different research groups with access to many samples of individuals across the full range of intelligence that have completed a large, diverse battery of cognitive tests, DNA analysis, and neuroimaging with structural and functional methods. We are just beginning to see such comprehensive projects, as noted in Chapters 2 and 4.
Super-memory cases are also of increasing interest. In Chapter 1, we mentioned Daniel Tammet’s recitation from memory of 22,514 digits of pi. According to the Guinness World Records, however, the record for reciting pi from memory is an amazing 67,890 digits. This record is held by a person (CL) who is not a savant. He uses mnemonic methods (i.e., memory tricks – see Textbox 6.2) that allow the storage and retrieval of large amounts of information. One fMRI study recruited several participants in the World Memory Championships and found several brain areas were activated when mnemonic procedures were used (Maguire et al., 2003). Unfortunately, each participant used a different mnemonic strategy, so the imaging results were not easily interpretable. At the age of 28 years, CL, the holder of the Guinness record for memorizing 67,890 digits of pi, was studied with fMRI while he used his strategy and a strategy designed by the researchers as a control condition (Yin et al., 2015). CL has many years of training on his mnemonic method, which the authors described this way: “CL used a digit–image mnemonic in studying and recalling lists of digits, namely associating 2-digit groups of ‘00’ to ‘99’ with images and generating vivid stories out of them.” An example of this method is created in Textbox 6.2. Eleven male graduate student controls were also scanned and tested in the same strategy conditions. According to the authors, the results suggested that CL relied on brain areas related to episodic memory rather than verbal rehearsal. The imaging results are actually quite complex and open to interpretation (Sigala, 2015).
Textbox 6.2: A memory trick
Here’s how you can train yourself to memorize a long string of numbers like pi just in case you decide to do so. Before memorizing the digits, create a list of words to represent 100 pairs of sequential digits. For example, if the sequence contained 0,0 that would be remembered as a dog. If the sequence contained 0,1 that would be remembered as a fish. Assign a word to the combination of 0,2 and another word for 0,3 and so on for 0,4 … 5,0, 5,1 … 9,9. The words can be, for example, animals, tools, famous historical figures or anything you choose. As you begin to memorize the long string of numbers convert sequential pairs to your pre-memorized list of 100 words and create a story linking each consecutive word. The more outrageous the story, the easier it is to remember. Let’s say this is your standard list of words for each two-digit combination (showing only 8 of 100 pairs):
00 dog
01 fish
02 Lincoln
03 hammer
… 29 robin
… 51 airplane
… 86 shoe
… 99 bank
Now here is a string of 18 numbers to memorize: 860229000299000151. Convert this string to pairs and then convert the pairs to your pre-assigned word for each pair: 86 is shoe; 02 is Lincoln; 29 is robin; 00 is dog; 02 is Lincoln; 99 is bank; 00 is dog; 01 is fish; 51 is airplane. So then you memorize a story you create that is rich in visual imagery, like: My shoe fits Lincoln and he kicks a robin that is eaten by a dog but Lincoln takes it to the bank where a dog is eating a fish on an airplane. With practice, this sentence is easier to remember and after you memorize it, you can convert the words back to two-digit pairs. It may seem quite awkward, but this kind of mnemonic strategy can be used effectively to remember many things, from numbers to names. It requires considerable practice and imagination, but some people get remarkably good at it and a few people are extraordinary. This is how CL memorized 67,890 digits of pi. Unlike some of the electrical or drug-enhancement techniques discussed in Chapter 5, this is one thing you can try at home. There is no evidence, however, that increasing your ability to memorize like this increases your intelligence (see Chapter 5). It is an open question as to whether people who learn to excel at memorizing using this method are people who already start with high intelligence scores.
Positron emission tomography (PET) was used in a similar study of a mental calculation prodigy (Pesenti et al., 2001). Mental calculators are exceptionally accurate and fast at solving complex calculations in their heads. Whereas some savants apparently have this ability without training, the person studied in this report, 26-year-old R. Gamm, is a healthy individual and not a savant. He had, however, “trained his memory for arithmetic facts and calculation algorithms several hours each day for about six years” starting at age 20. For example, he could calculate two-digit numbers to various powers (e.g. 995 equals 9,509,900,499 or 539 equals 3,299,763,591,802,133). He could also do roots, sines, divisions of prime numbers, and apply an algorithm to perform calendar calculations to name the day of the week for any date (another ability found in savants). Gamm was compared to six non-expert male students scanned as controls performing the same tasks during PET determination of regional blood flow. PET scans were obtained during a calculation task and a memory retrieval task. The results showed brain activations in several areas common to both Gamm and the controls, but Gamm also showed unique activations when the two task conditions were contrasted. Gamm showed more activation in medial frontal and the parahippocampal gyri, the upper part of the anterior cingulate gyrus, the occipito-temporal junction in the right hemisphere and the left paracentral lobule (see where these areas are in Figure 6.1). The authors concluded that, “… calculation expertise was not due to increased activity of processes that exist in non-experts; rather, the expert and the non-experts used different brain areas for calculation. We found that the expert could switch between short-term effort requiring storage strategies and highly efficient episodic memory encoding and retrieval, a process that was sustained by right prefrontal and medial temporal areas.” In other words, Gamm’s brain worked differently.
Figure 6.1 PET scans of an expert memory champion performing complex mental calculations compared to six non-expert controls. Brain areas uniquely activated in the expert are shown in green; areas activated both in the expert and non-experts are shown in red. Bar graphs show activations in each area for each person (red bar is the expert).
Reprinted with permission, Pesenti et al. (2001).
Imaging such rare individuals with exceptional mental ability achieved by training
may provide insights about the effects of intensive strategy training over years on brain networks, or insights about unusual brain connections that apparently result by chance, or unknown factors in the case of savants. There is no indication that either CL or R. Gamm showed an increase in g related to their respective intensive memory training.
6.3 Bridging Human and Animal Research with New Tools Neuron by Neuron
On the much smaller spatial scales of neurons and synapses, intelligence is not a major focus of interest for most neuroscience researchers. There are some attempts to relate neurotransmitters and other aspects of synaptic function to intelligence in molecular genetic studies, as we noted in Chapters 2 and 4. Many questions remain ripe for examination. For example, does the number or type of mitochondria inside neurons (from any particular brain area) have any relationship to individual differences in the g-factor or other mental abilities? An older postmortem human study suggested a relationship between the complexity of dendrites and education level (indirect measure of intelligence) (Jacobs et al., 1993), but the direction of the relationship could go either way and replication is required. There are many possibilities to study intelligence on this level, especially if technology ever advances to the point where non-invasive measurements of single neurons and synapses can be made in humans.
Until such a time, animal studies provide some intriguing observations that suggest a tentative bridge to human studies. A systematic lesion study in rats, for example, found several discrete brain areas were related to general problem-solving ability because lesions to those areas degraded performance on several different problem-solving tasks (Thompson et al., 1990). Lesions to other areas degraded performance only on specific tasks. The areas implicated in this study were compared to early PET studies of reasoning in humans, but showed only limited overlap (Haier et al., 1993). Nonetheless, this combination of problem-solving tasks and lesions provided an animal model for brain/intelligence relationships that expanded the pioneering work of Karl Lashley (1964) and indicated that the g-factor is not unique to humans. Studies of genetically diverse (outbred) mice learning a variety of tasks also indicate a g-factor. The results from one study of mice sound strikingly like those from human studies: “Indicative of a common source of variance, positive correlations were found between individuals’ performance on all tasks. When tested on multiple test batteries, the overall performance ranks of individuals were found to be highly reliable and were ‘normally’ distributed. Factor analysis of learning performance variables determined that a single factor accounted for 38% of the total variance across animals. Animals’ levels of native activity and body weights accounted for little of the variability in learning, although animals’ propensity for exploration loaded strongly (and was positively correlated) with learning abilities. These results indicate that diverse learning abilities of laboratory mice are influenced by a common source of variance and, moreover, that the general learning abilities of individual mice can be specified relative to a sample of peers” (Matzel et al., 2003). They also demonstrate the importance of an individual differences approach, even in mice (Sauce & Matzel, 2013).
Continuing this line of research, Matzel and Kolata summarized human imaging studies of memory/intelligence and mice experiments that tested causal relationships between aspects of selective attention, working memory and general cognitive ability (Matzel & Kolata, 2010). They concluded that the data suggested “that common brain structures (e.g., prefrontal cortex) mediate the efficacy of selective attention and the performance of individuals on intelligence test batteries. In total, this evidence suggests an evolutionary conservation of the processes that co-vary with and/or regulate ‘intelligence’ and provides a framework for promoting these abilities in both young and old animals.” Having such potent animal models of intelligence can help drive future neuroscience experiments, especially moving down the spatial and temporal scales from accumulated brain activity in specific areas to more precise measurements in neurons and synapses using methods not applicable to humans. There is some suggestion, for example, that physical and mental training in mice may increase the number of neurons and their survivability in specific brain areas (Curlik et al., 2013; Curlik & Shors, 2013). There is also some suggestion in mice that the signaling efficiency of the dopamine D1 receptor in the prefrontal cortex may relate to both memory tasks and intelligence tests (Kolata et al., 2010; Matzel et al., 2013). It is too early to evaluate whether these findings represent a weight of evidence, but these studies demonstrate how an animal model of intelligence will help direct neuroscience progress down to the level of neurons and synapses.
Another illuminating example is the use of fluorescent proteins that literally light up neurons and synapses. The first fluorescent protein was discovered in jellyfish decades ago and that discovery has evolved into amazing techniques that create new fluorescent proteins and remarkable ways to introduce them into cells. Once inside a neuron, fluorescent proteins can track electrical activity and map neural circuits in the brain. Different fluorescent proteins attach to different neurochemicals and produce different colors. This means that the distribution of individual neurotransmitters can be mapped. In fact, individual neurons can be made to have a unique color so individual neuron pathways and their neurochemical signals can be mapped. Fluorescent studies of intelligence in mice would be fascinating. Doogie meets Mickey in a new Fantasia movie.
In the previous chapter, we briefly introduced one photo-neuro-modulation method that used red laser light to activate or deactivate neurons. Newer optogenetic and chemogenetic methods are more specific and based on modifying synaptic receptors so neurons react to special light-sensitive chemicals. Both methods have been used to modify mouse behavior by turning on light. In the process, experimental studies reveal neuro-circuits involved in complex behaviors and suggest ways to modify them.
The optogenetic method basically works like this. Normally, neurons fire when they receive a brief electrical pulse across synapses from neighboring neurons. This pulse changes the neurochemistry of the receiving neuron to create another pulse that travels to neighboring neurons in the circuit. The key neurochemical change involves proteins within the neuron. The electrical pulse stimulates the protein to create a new pulse to fire the next neurons in the circuit, and this cascade of firing continues until inhibitory signals diminish or stop the firing. Optogenetic techniques create light-sensitive proteins in specific populations of neurons. These neuron clusters can be induced to fire by applying light in controlled experiments. Light is delivered directly into neurons using hair-thin fiber-optic thread after light-sensitive proteins have been introduced genetically into the neurons of interest. In a mouse model of depression, for example, light stimulation of neurons in the medial frontal cortex relieves symptoms (Covington et al., 2010). Symptoms of cocaine addiction in mice can be reversed by light stimulation to neurons that project to the nucleus accumbens (Pascoli et al., 2012). Aggressive or sexual behavior in mice can be activated when a burst of light stimulates different neurons in the hypothalamus (Anderson, 2012). Optogenetic methods can be combined with the CRISPR-Cas9 method of gene editing (described in Chapter 5) so specific gene expression (turning genes on and off) can be targeted with light (Nihongaki et al., 2015). This incredible field is growing rapidly and there are many examples of experiments that could eventually lead to therapies for brain disorders, and perhaps to enhancement of mental abilities (Aston-Jones & Deisseroth, 2013; Wolff et al., 2014). It sounds like science fiction, but it is happening now in a laboratory near you. Screenwriters, pay attention.
The chemogenetic method is a complementary approach for turning neurons on and off. This technique is based on creating “designer receptors exclusively activated by designer drugs,” known by the acronym DREADD (Urban & Roth, 2015); how do they find these names? Recently, researchers developed a variation of DREADD that allowed neurons to be turned on and off, rather than the previous limitation of on or off techniques (Vardy et al., 2015).
This allowed these researchers to turn hunger and activity levels in mice on or off for periods of time longer than can be done with optogenetic methods.
Neuroimaging methods described in Chapters 3 and 4 give researchers a view of the brain like the view of a city from a high-flying airplane; a unique and informative view not possible before the invention of the airplane. These new neuroscience techniques give researchers experimental control over individual neurons. This is like an aerial view that allows seeing individual cars on a city street and possibly who is in the car and how fast their heart is beating. We can only imagine further refinements, new DREADDs, and new experiments. There is breathtaking potential for elucidating intelligence brain circuits if these techniques are applied to animal models of intelligence like those described in this section by Matzel and colleagues. If such methods are available in humans, the potential for neuroscience/intelligence research staggers the imagination. Ready to change your major or thesis topic?