Working memory training: Improving intelligence – Changing brain activity

Abstract

The main objectives of the study were: to investigate whether training on working memory (WM) could improve fluid intelligence, and to investigate the effects WM training had on neuroelectric (electroencephalography – EEG) and hemodynamic (near-infrared spectroscopy – NIRS) patterns of brain activity. In a parallel group experimental design, respondents of the working memory group after 30 h of training significantly increased performance on all tests of fluid intelligence. By contrast, respondents of the active control group (participating in a 30-h communication training course) showed no improvements in performance. The influence of WM training on patterns of neuroelectric brain activity was most pronounced in the theta and alpha bands. Theta and lower-1 alpha band synchronization was accompanied by increased lower-2 and upper alpha desynchronization. The hemodynamic patterns of brain activity after the training changed from higher right hemispheric activation to a balanced activity of both frontal areas. The neuroelectric as well as hemodynamic patterns of brain activity suggest that the training influenced WM maintenance functions as well as processes directed by the central executive. The changes in upper alpha band desynchronization could further indicate that processes related to long term memory were also influenced.

Introduction

Attempts to improve intelligence are by no means new in psychology. The main objective is to improve fundamental processes that form the basis of intelligent behavior and in that way increase general intelligence (G), or fluid intelligence (Gf). It is of course very easy to increase test performance by simply practising the tests themselves, or by practising similar tasks. However, since Jensen, 1969, Jensen, 1981 claim that interventions aiming to improve intelligence resulted in only very little if any success at all, only sporadic attempts have been made to investigate interventions that could increase ability. To mention just one, the highly controversial Mozart effect. College students after 10 min of listening to Mozart’s Sonata (K. 448) had Stanford–Binet spatial subtest IQ scores 8–9 points higher than students who had listened to a relaxation tape or listened to nothing. The IQ effects did not persist beyond the 10–15 min testing session. (Rauscher, Shaw, & Ky, 1993, for a review see, Pietschnig, Voracek, & Formann, 2010).

Recently, the debate on whether certain interventions can increase ability has once more gained popularity. The discussion has been triggered by the study of Jaeggi, Buschkuehl, Jonides, and Perrig (2008), showing that working memory (WM) training can increase fluid intelligence. Jaeggi et al. (2008) have shown that an increase in fluid intelligence can be obtained by training on problems that, at least superficially, do not resemble those on the ability tests. They could further show that more training leads to greater IQ gains, which were present across the whole spectrum of abilities, although larger toward the lower end of the spectrum. Despite positive comments like: “Their study therefore seems, in some measure, to resolve the debate over whether fluid intelligence is, in at least some meaningful measure, trainable.” (Sternberg, 2008, p. 6791), the study and its design have been also criticized (Moody, 2009, Sternberg, 2008). Sternberg (2008), not calling into question the obtained results, stressed eight limitations of the findings reported by Jaeggi et al. (2008). Among the major ones were the lack of a placebo (active) control group, the use of just one training task, and only one measure of the dependent variable Gf. Even more strict was the criticism put forward by Moody (2009). The main objection raised was that different tests were used for the so-called control group (receiving just 8 days of training), and the experimental group, where individuals were tested with an alternative test with a time restriction that may have biased results. Such a time restriction made it impossible for respondents to solve the more demanding tasks. Since the whole weight of Jaeggi’s (Jaeggi et al., 2008) conclusions rests upon the validity of the measure of fluid intelligence, in Moody’s (2009) opinion, this brings into question the results and inferences reported in the study.

Buschkuehl and Jaeggi (2010), in a review of 11 studies aimed at improving intelligence, divided the interventions used to influence ability into approaches that were focused on training of WM and executive functions, and interventions which entailed other approaches – video games and other cognitively stimulating activities, like music, or supplementing participants with creatine. Two conclusions have been put forward: First, most of the studies, are heterogeneous on several dimensions and have certain methodological shortcomings, yet most of them reported significant improvements in measures of ability. Second, most numerous were the attempts to improve intelligence by WM training tasks (Klingberg, 2010). This seems reasonable, as there is a strong link between WM and intelligence (Colom, Abad, Quiroga, Chun, & Flores-Mendoza, 2008). Further, there exist well elaborated models of WM, like the multi component model of Baddeley (for a review see Repovš & Baddeley, 2006), or the embedded processes model proposed by Cowan (1999). The models do differ, but they define WM function as the combination of short-term storage and some sort of processing components. It is further worth mentioning that in a recent study by Colom et al. (2008) it could be shown that short-term storage largely accounted for the relationship between working memory and intelligence, and that processing components, like mental speed, updating, and the control of attention had a negligible, or no relation to intelligence.

Morrison and Chein (2011) in a review of WM training studies divided the training approaches into strategy training (intended to promote the use of domain-specific strategies to remember increasing amounts of information), and core training (repetition of demanding WM tasks targeting general WM mechanisms). Their conclusion was that core WM training produced more far-reaching transfer effects. However, they also discussed several limitations that should be addressed. Among the most important were: no control of motivational (expectancy) and repetition effects, and a lack of consistency in experimental methods. A similar conclusion, namely that inadequate control and ineffective measurement of the cognitive abilities of interest cloud the interpretation of the current training literature, was put forward by Shipstead, Redick, and Engle (2010).

A much less favorable view with respect to the trainability of ability was provided by Owen et al. (2010). In their large-scale study, 11,430 individuals participated in a 6-week online training of different cognitive tasks designed to improve reasoning, memory, planning, visuospatial skills and attention. The findings led the authors to the conclusion that, although improvements were observed in every one of the cognitive tasks for which were trained, no evidence was found for transfer effects to untrained tasks, even when those tasks were cognitively closely related, or to any general improvement in the level of cognitive functioning.

The second strand of research, which is fostering the debate on the possibility of increasing the level of intelligence, comes from the broad area of neuroscience. Until recently neurologists were convinced that neural plasticity is present only in childhood. Plasticity of the nervous system denotes developmental changes in synaptic density and synaptic pruning, and plays the key role in cell loss, and the growth and myelination of white matter (Craik, 2006). There is also evidence that there is some plasticity and fine-tuning that continues across the lifespan. Maguire et al. (2000) found that in London taxi-drivers the posterior region of the hippocampus is much larger than in the rest of the population, whereas the front region is much smaller. One important role of the hippocampus is to facilitate spatial memory in the form of navigation. Similarly, an enlargement of the auditory cortex (25%) in highly skilled musicians, compared with people who never played an instrument was reported by Pantev et al. (2003). That such changes can be rather rapid was shown by Pascual-Leone (2001). Even the amount of five days practising a five-finger piano exercise enlarged areas of the brain responsible for finger movements. On the other hand, when practising stops, the brain tends to return to its normal size. This was shown by a study where people learned to juggle for 3 months. After training, an increase in size in the midtemporal area and the left posterior intraparietal sulcus (areas responsible for visual motion information) was observed. Nonetheless, after 3 months of no practice, these areas returned to their previous size (Draganski et al., 2004, Driemeyer et al., 2008).

In the light of these findings, one could expect that training aimed to increase intelligence, would be also reflected in brain functioning. Further support for this view was provided by several neurofeedback studies. The study by Keizer, Verment, and Hommel (2010) has shown that neurofeedback in the gamma-band (36–44 Hz) could improve episodic retrieving. In another study by (Zoefel, Huster, & Herrmann, 2011), neurofeedback training of the upper alpha frequency band improved cognitive performance on a mental rotation task. Egner and Gruzelier (2003) could further show that learning to progressively raise theta (5–8 Hz) over alpha (8–11 Hz) band amplitudes significantly enhanced music performance. Several additional studies conducted by the author (Gruzelier, 2009) showed the positive influence of theta/alpha neurofeedback on creativity and ballroom dancing. Furthermore, the study by Surmeli and Ertem (2010) showed that quantitative guided neurofeedback could significantly improve intelligence in a group of mentally retarded patients.

Research into the trainability of intelligence and related changes in brain activity could also improve our understanding of the relationship between the psychometric construct of G or Gf and brain functioning. There exist numerous findings for the intelligence–brain relationship, but few theories. In a recent overview on research pertinent to the relationship between psychometrically determined intelligence and functional characteristics of brain activation observed during cognitive task performance, Neubauer and Fink (2009) concluded that most of the reviewed studies have demonstrated a negative correlation between brain activity under cognitive load and intelligence. The explanation of these findings was an efficiency theory – the non-use of many brain areas irrelevant for task performance, as well as the more focused use of specific task-relevant areas in individuals with high IQ’s (Haier, 1993). Some studies have shown a specific topographic pattern of differences related to the level of intelligence. High-ability subjects made relatively greater use of parietal regions, whereas low-ability subjects relied more exclusively on frontal regions (Gevins and Smith, 2000, Jaušovec and Jaušovec, 2004). It was further reported that highly intelligent subjects displayed more brain activity in the early stages of task performance, while average individuals showed a reverse pattern. This temporal distribution of brain activity suggested that cognitive processes in highly intelligent individuals are faster than in average intelligent individuals (Jaušovec & Jaušovec, 2004). A further finding was also that neural efficiency seems to be corroborated mostly when participants work on tasks of low to medium difficulty or complexity (Neubauer & Fink, 2009). In the study by Doppelmayr, Klimesch, Hödlmoser, Sauseng, and Gruber (2005), the expected findings of a negative relation between brain activation and intelligence emerged solely for the easier items of the Raven test, whereas a tendency in the opposite direction was observed for the difficult ones. It was further shown that less intelligent individuals displayed a decrease in activation from easy to difficult tasks, whereas the opposite was true for the brighter participants. It is likely that individuals with low IQ’s did not even try on the harder problems, which could explain their lower activation levels compared to those with high IQ’s. It could be further assumed that individuals with low IQ’s have to work harder on easy problems than do individuals with higher IQ’s.

Research has also focused on structural correlates of human intelligence, attempting to answer the question: “Where in the brain is intelligence?” This body of evidence has recently been synthesized by Jung and Haier (2007) in the form of their so-called ‘parieto-frontal integration’ (P-FIT) model of intelligence. In reviewing 37 neuroimaging studies mostly on structural correlates of intelligence they tried to answer the question of how the anatomical aspects of gray matter and white matter relate topographically to intelligence. The P-FIT model by Jung and Haier (2007) suggested – contrary to the assumption of Duncan (2001) that intelligence is localized in the pre-frontal cortex – that beside frontal areas of the brain also the temporal and occipital lobes are critical in early processing of sensory information which is then fed forward to the parietal cortex, wherein abstraction, and elaboration emerge. These processes are dependent upon the fidelity of underlying white matter necessary to facilitate rapid and error-free transmission of data from posterior to frontal brain regions. The main problem with the P-FIT theory is that only a very small number of discrete brain areas approach 50% of convergence across published studies employing the same neuroimaging technique (Colom et al., 2009, Haier et al., 2009). When different test batteries were used to derive G, this changed also the brain areas related to G.

The aim of the present study was to investigate whether training of WM functions (short-term storage and processing components like control of attention and executive functioning) can improve performance on tests of fluid intelligence. A second objective was to investigate the influence of WM training on brain activity. In order to gain deeper insight into brain activity under cognitive load a multi-modal brain imaging approach was used, combining electroencephalographic (EEG) methodology (based on electrophysiological principles) with near infrared spectroscopy (NIRS) imaging technique (based on hemodynamic principles). Such a multi-modal approach compensated for the poor spatial resolution of EEG and allowed for a more reliable testing of the hypotheses described below.

Further, the experiment was designed to eliminate some of the criticisms (Moody, 2009, Sternberg, 2008) of Jaeggi’s study (Jaeggi et al., 2008), and integrate suggestions for improvements put forward in recent reviews of research (Buschkuehl and Jaeggi, 2010, Klingberg, 2010, Morrison and Chein, 2011) on WM training (e.g., an active control group was included, different types of training tasks as well as different tests of fluid intelligence were used). Given the exploratory nature of the study our expectations were rather general. It was expected that the training would significantly increase test performance of respondents in the experimental group. It was further expected that these differences would be reflected in brain functioning mainly in the frontal and parietal brain areas of respondents and that no such differences would be observed in respondents of the active control group. In considering the relatedness of EEG frequency bands with different cognitive functions it was expected that the main test–retest differences in the experimental group would be observed in the theta and lower-1 alpha bands. Several studies (Klimesch, 1999) have shown that episodic and working memory processes were reflected in theta band synchronization (ERS), and that attentional processes were related to lower alpha band desynchronization (ERD), while on the other hand, semantic or long term memory (LTM) processes were associated with upper alpha band desynchronization (Klimesch et al., 1994).