Annual Research Review: Educational neuroscience: progress and prospects

In this section, we consider a recent funding round for projects in educational neuroscience to illustrate which types of projects are being invested in. We then consider the main areas of established research findings, in areas of brain health, cognition and core skills, adolescence, executive function, social cognition, emotion and developmental disorders, as well as the quest for techniques that will yield general enhancement of cognition.

Example projects

In the United Kingdom in 2014, the Education Endowment Foundation (a charity with the mission to improve the educational attainment of the poorest pupils in English schools) and the Wellcome Trust launched an “Education and Neuroscience scheme”. The goal was to provide funding for collaborative projects between educators and neuroscientists to develop evidence-based interventions for use in the classroom (e.g. Howard-Jones, 2014b). The six research projects the scheme funded, some of which are still ongoing, give a snapshot of the make-up of current research initiatives in educational neuroscience. One project examined sleep patterns in adolescents. It trialled later school start times, along with a sleep education programme, to assess their impact on teenagers' educational achievement (Espie et al., 2016; Kelley, Lockley, Foster, & Kelley, 2015). Another project investigated the effect of medium to high cardiovascular activity on academic attainment, using brain imaging to investigate the mechanistic basis of any observed correlation (see Johansen-Berg & Duzel, 2016).

The next project investigated the potential of inhibitory control training on primary age children's learning of mathematics and science, based on neuroscience evidence that conceptual learning in these domains requires suppression of pre-existing beliefs, particularly for counterintuitive concepts (Mareschal, 2016) – for example, learning the world is round when it seems flat. The fourth project assessed the effectiveness of ‘spaced learning’, a particular regime of repeating a unit of work three times interspersed with alternative activities (Kelley & Whatson, 2013). The fifth project drew on the neuroscience of reward-based learning, evaluating the effectiveness of uncertain reward on attainment in science (Howard-Jones & Jay, 2016). The final project investigated the potential of a computer game to improve reading via developing phonological awareness through ‘rhyme analogy’ (Kyle, Kujala, Richardson, Lyytinen, & Goswami, 2013).

Brain health

The first two projects illustrate the direct route from neuroscience to education. The direct route seeks to improve educational outcomes by enhancing the operation of the brain as a biological organ. This can be viewed as part of a broader approach of improving ‘brain health’ or of ‘brain optimisation’, concepts that extend beyond the school walls to the lifestyles of children and adults at home. Brain health targets physical fitness, diet, sleep and stress reduction through relaxation or meditation techniques and considers environmental factors such as air and noise pollution. By addressing these factors, the child should arrive in the classroom in the optimal state to learn. For example, restricting adolescent sleep time to just 5 or 6 hours a night for several nights has a marked impact on cognitive performance, most consistently on working memory ability (Jiang et al., 2011; Lo, Ong, Leong, Gooley, & Chee, 2016), even in the absence of subjective feelings of sleepiness. Conversely, increasing the regularity and intensity of physical exercise during the school day typically improves both academic performance and cognitive test scores (see Ruiz-Ariza, Grao-Cruces, Marques de Loureiro, & Martínez-López, 2017). The role of relaxation and meditation practice on learning outcomes is also an area of active research. Meditation putatively serves not just to reduce stress but perhaps also to improve executive function skills and emotion regulation by helping children actively engage and direct their own attention. Although published intervention studies have not yet looked at the impact of meditation on academic performance, research indicates a positive impact on prosocial psychosocial attributes such as classroom behaviour and a decrease in reported and observed psychopathology in children (see Felver, Celis-de Hoyos, Tezanos, & Singh, 2016).

Epidemiological studies have suggested that highly polluted environments are associated with delays in cognitive development as shown, for example, by a recent prospective study of 7–10-year-old children (n = 2,715) in Barcelona (Sunyer et al., 2015). Proposals have been advanced for the biological basis of this effect, such as translocation of inhaled ultrafine particles into the brain, with associated induction of a low-level chronic neuroinflammation and oxidative stress, resulting in glial activation and white matter injury (Donaldson et al., 2005). Notably, in focusing on biology, the direct route makes use of one particular neuroscience method, animal models. It can be difficult to infer specific causal mechanisms from naturalistic data sets containing complex sets of correlations between environmental factors and educational outcomes. In animal models, a factor like air pollution can be randomly assigned and physiological changes can be observed in neurons that support learning. Establishing causal links between pollution and brain damage provides a much more compelling argument for stricter limits on pollution near schools. Without this kind of evidence, naturally existing correlations between pollution and poverty might make it difficult to offer policy recommendations.

Cognition and core skills

The other four example projects illustrate the indirect route. They build on developmental cognitive neuroscience theories to propose and evaluate novel learning activities and their transfer to educational achievement. Respectively, the projects draw on theoretical work on concept formation and its relation to inhibitory control (that is, the capacity to voluntarily regulate strong or automatic behavioural responses), long-term memory formation and its relation to long-term potentiation, reward-based learning and dopaminergic modulation of hippocampal learning, and brain sensitivity to print generated when children learn letter to speech sound correspondences at different levels of granularity.

One of the major strategies in the indirect route has been to use neuroscience evidence to help identify the core cognitive skills required for acquisition of domains such as literacy and numeracy, with the potential to identify the causes of deficits (for example, in one or more of the core skills) and training techniques that would target these skills. Work in numeracy has used imaging methods to understand the relationship between the learning and the respective core skills of the perception of number symbols (in fusiform gyrus in occipitotemporal lobes), representation of numerosity and manipulations of quantities (respectively, in intraparietal sulcus and the angular gyrus within the parietal lobe), spatial abilities (in the parietal lobe) and concepts, principles and procedures (involving prefrontal cortex; Butterworth & Varma, 2013). To illustrate the kinds of questions currently under consideration, a theoretical debate continues on whether the approximate non-symbolic representations of number (e.g. dot arrays) form the basis for symbolic number processing (Merkley & Ansari, 2016), or whether more abstract representations are necessary to forge the connections between language, spatial representations and perceptual attention systems for tracking small numbers of objects. The identification of a set of core underlying systems has raised the possibility of training the parietal ‘approximate number systems’ for processing quantity (Budgen, DeWind, & Brannon, 2016), visuospatial working memory for counting and subitising (Menon, 2016) and spatial processing for using material such as maps, diagrams and graphs (Newcombe, 2016).

Research in other academic disciplines is less advanced. Science comprises a diversity of skills, including concept formation and reasoning. The recent focus has been on how the acquisition of scientific concepts can be integrated with prior intuitive concepts, for instance about the nature of the physical world (e.g. the Earth seems flat, heavy objects seem to fall faster than light ones) and on how intuitive reasoning can be extended to hypothesis testing (see Mareschal, 2016; Fugelsang & Mareschal, 2013). There is, relatively speaking, less topic-specific educational neuroscience pertaining to topics such as history, geography, art and sports (though perhaps music is an exception, see e.g. Koelsch, 2012; Peretz & Zatorre, 2003,).

Adolescence

There has been a significant amount of work in educational neuroscience into adolescence (see, e.g. Blakemore, 2018; Fuhrmann, Knoll, & Blakemore, 2015). Notably, the focus on adolescence was sparked by neuroscience evidence that the brain exhibits an extended developmental trajectory continuing through adolescence and into early adulthood (e.g. Dumontheil, 2016, for review). For example, although it was thought intelligence stabilised by 10 years of age, justifying educational selection according to ability at aged 11, Ramsden et al. (2011) demonstrated that an individual's IQ score, measured with standardised verbal and non-verbal tests, can fluctuate across adolescence. Importantly, changes in brain structure closely mirrored the changes in IQ, indicating that fluctuating IQ scores were not simply the result of errors in test administration, design, coding or participant motivation, but represented real changes in ability. Moreover, when Knoll et al. (2016) compared training of relational reasoning against a control task of face discrimination in a sample of over 600 individuals from 11 to 33 years, they found that gains in relational reasoning were stronger in older adolescence and young adulthood than in early adolescence. Meanwhile, training on the perceptual task of face discrimination yielded no reliable gains in any age group. For the more complex task, learning ability was greater in the older groups.

The extended developmental trajectory in adolescence occurs unevenly across the brain, with temporal and frontal areas the last to mature. It has been argued that this aspect of brain development, possibly driven by pubertal hormonal changes (Piekarski, Boivin, & Wilbrecht, 2017), might offer insights into late developing skills in adolescence, such as (responsible) decision-making and perspective taking (e.g. Mills, Goddings, Clasen, Giedd, & Blakemore, 2014). In addition, there may be specific changes linked to puberty that alter reward-based processing in adolescence and vulnerability to peer influence, a particular characteristic of the teenage years (Crone & Dahl, 2012; Van Hoorn, Fuligni, Crone, & Galvan, 2016).

Executive functions, social cognition and emotion

While branches of educational neuroscience address cognitive skills relevant to particular academic domains, there is also work that considers cognitive processes that impact all domains. Perhaps the most studied are ‘executive functions’, a set of related processes involving cognitive control and flexibility (Diamond, 2013). Executive function skills are strongly predictive of academic achievement, for example, predicting up to a third of the variance in maths and reading scores (Best, Miller, & Naglieri, 2011; Samuels, Tournaki, Blackman, & Zilinski, 2016). Cognitive neuroscience research continues to investigate how the improvement of executive functions with age is constrained by the development of relevant prefrontal brain regions (Kharitonova, Martin, Gabrieli, & Sheridan, 2013). This is relevant not only for the role of executive functions in subsequent academic achievement but also, at a younger age, for school readiness – how well a child can follow instructions, engage and take part in the learning environment of the classroom (Blair, 2016; De Haan, 2013). Executive function skills are potentially trainable (Diamond & Ling, 2016) and, importantly, are a remediable weakness in children raised in deprived home environments (Neville et al., 2013).

Educational neuroscience research on cognitive control also extends to the regulation of emotion. It is sometimes necessary for children to regulate emotions in educational contexts, and this is a skill that develops across early childhood (McRae, 2016). Here again, researchers have investigated how emotion regulation is associated with the development of a set of prefrontal regions involved in executive functions and their connections to limbic systems (Martin & Ochsner, 2016). Emotion has been considered in relation to specific academic domains, particularly with respect to maths anxiety (Beilock & Maloney, 2015; Chang & Beilock, 2016). However, emotion has a wider relevance to the social classroom environment and the relationships between students and between students and teachers (Immordino-Yang & Gotlieb, 2018). Research into adolescence has focused on motivation and psychological well-being, with heightened risk of mental health problems in the teenage years (Jones, 2013). Finally, work in educational neuroscience has begun characterising the development of the social brain (Blakemore, Kadosh, Sebastian, Grossman & Johnson, 2013). This includes mechanisms underlying abilities such as gaze processing, joint attention, face processing, action observation and reasoning about other people's mental states; for adolescence, one particularly important area is peer acceptance and rejection (Van Hoorn et al., 2016).

In contrast to work on executive functioning, where specific training interventions are common, research on social and emotional processing is not as advanced in translating insights from cognitive neuroscience into classroom implications, despite recognition of the important role of schools in pupils' social and emotional development (UK Department for Education, 2007).

Developmental disorders

Bruer (2013) argued that the most likely payoff for educational neuroscience will be in the area of special education, addressing the needs of children suffering from developmental language disorder, dyslexia, dyscalculia, attention deficit and related executive function disorders, and social and emotional disorders. Much research effort in cognitive neuroscience has been dedicated to these areas (e.g. Rinehart, Bradshaw, & Enticott, 2017). The goal is to identify the underlying mechanistic causes of the deficit, be they in brain areas that are under developed, brain areas that are under-activated, or brain areas that are poorly connected; and to use these data to inform cognitive theories. The search for mechanistic explanations contrasts with and complements research approaches in educational psychology that address issues of exclusion/inclusion of children with special educational needs (SEN) from mainstream education and the psychosocial perspective of families with children with disabilities (see, e.g. Woolfson, 2011). However, a mechanistic understanding of the cause of a developmental deficit does not necessarily provide a straightforward pathway to an intervention to remediate the deficit – particularly when therapeutic interventions may exploit compensatory strategies as much as seek to alleviate the mechanistic deficit (Bowers, 2016a). Indeed, our current understanding of the mechanistic basis of compensatory strategies lags behind that of causes of deficits themselves (Thomas et al., 2018). Insights into the causes of deficits must be transformed by pedagogical principles into interventions, which in turn must be evaluated for their effectiveness by behavioural trials in educational contexts (Howard-Jones et al., 2016).

One interesting case study to watch is the emergence of a new mechanistic theory of developmental dyslexia along with new avenues for intervention. The ‘temporal sampling’ theory of dyslexia (Goswami, 2017), motivated largely by findings from electrophysiology, proposes that the functional deficit in dyslexia is a difficulty using the temporal structure of speech to scaffold the perception of syllables and sentence prosody. To complement experimental work on this question, it has been found that highlighting auditory rhythmic information in non-speech and speech stimuli is as effective at improving phonological awareness in children with dyslexia as directly targeting phonology (Thompson, Leong, & Goswami, 2013). Interventions for children with dyslexia may soon, therefore, focus on music and rhythm rather than reading (Flaugnacco et al., 2015; Goswami, 2017).

To date, there are good examples of neuroscience contributing to the understanding of the causes of developmental deficits, such as atypical development of parietal systems supporting number processing in dyscalculia (Butterworth, Varma, & Laurillard, 2011) or of the attention and executive function networks in prefrontal cortex in attention deficit hyperactivity disorder (Shaw et al., 2013). Studies of the brain mechanisms of learning may also allow better prediction of who will benefit from different intervention techniques. This has been shown in the case of children with dyslexia in whom brain responses during a reading task predicted response to intervention when behavioural measures held no predictive value (Hoeft et al., 2011). Neuroscience has also had a wider and subtler impact on the debate surrounding the utility of defining specific categories of learning disability and allocating diagnoses to individual children according to these categories. Investigations at both brain level (Peters, Bulthé, Daniels, Op de Beeck, & De Smedt, 2018) and genetic level (e.g. Kovas, Haworth, Dale, & Plomin, 2007) suggest that, genetic syndromes aside, most education-relevant developmental deficits lie on a continuum with variation in abilities in the mainstream population. Moreover, little evidence has accrued for special teaching techniques applicable to specific development disorders that differ from techniques used with mainstream children, that is, for an SEN-specific pedagogy (Davis & Florian, 2004). Educational neuroscience, therefore, contributes to the debate on the pros and cons of diagnosis and labelling. For example, there are many symptoms in common across struggling learners diagnosed with different categories of disorder, while symptom variability is very high for children with the same diagnosis (e.g. Gathercole et al., 2016). Such insights are to be pitted against potential pragmatic benefits of allocating diagnoses in helping to attract resources for, and guide the choice of, effective interventions.

The pursuit of techniques to produce general benefits for cognition

As we have seen, cognitive training tends to produce only ‘near transfer’, that is, performance improvements to the task trained on or similar tasks, rather than more broadly across cognition (Sala & Gobet, 2017). Nevertheless, there remains a desire within educational neuroscience to find elusive activities that bring more widespread benefits to cognition. There is enthusiastic engagement from teachers/public/media when such possibilities are raised, enthusiasm which is sometimes in advance of the actual science. This trend for seeking general benefits can also be found within the commercial sector in so-called ‘brain training’ products – though to date, commercial products have not yet achieved the advertised far transfer (Simons et al., 2016).

Within educational neuroscience, a number of possibilities are currently under investigation, among them executive function training (Diamond & Ling, 2016), mindfulness training (see Felver et al., 2016), playing chess (Sala & Gobet, 2016), action video game playing (Bediou et al., 2018), learning a musical instrument or a second language (Moreno, Lee, Janus, & Bialystok, 2015), sleep (Sharman, Illingworth, & Harvey, in press) and aerobic fitness training (see Ruiz-Ariza et al., 2017). The jury is still out on many of these activities. Research is rendered more difficult on the one hand by the frequent lack of random allocation of participants to conditions, allowing for the possibility of confounds (e.g. that in some populations, bilinguals might also have higher education or SES levels; that those who successfully learn musical instruments may be more intelligent or dedicated to practise; that those who persistently play action video games may tend to have faster sensorimotor responses); and on the other hand, by the challenge of designing intervention studies to achieve random allocation. One recent review of the cognitive benefits of action video game playing suggested that intervention studies with random allocation produced effect sizes of around a third of the size of those observed in correlational studies without random allocation (Altarelli, Green, & Bavelier, in press). This suggests both possible effects of the training, and also that there are pre-existing differences between the sorts of young adults who frequently play action video games compared to those who do not. After promises of broad benefits, detailed research has sometimes revealed that transfer is not as far as first anticipated [e.g. for action video games, the main transfer is to selective visual and auditory attention (Altarelli et al., in press); working memory training improves performance on other working memory tasks but not other components of executive functioning, such as cognitive flexibility or inhibition (Diamond & Ling, 2016; Melby-Lervåg, Redick, & Hulme, 2016)] In any event, for each of these putative generally beneficial activities, it is desirable for investigators to propose and evaluate the cognitive and brain structures that mediate the transfer from training task to other cognitive skills. The less plausible the underlying mechanistic basis for the transfer, the more critically the published evidence in favour of the transfer must be examined.