Annual Research Review: Educational neuroscience: progress and prospects

Before we consider areas where educational neuroscience has made most progress, it is worth pointing out why the goal of the field is a very challenging one. First, the way the brain learns is complex. Second, learning is only one part of education. Third, society's goals for education are not necessarily clear. And fourth, even for psychology, successful translation from science to educational practice has proved difficult. We consider these points in more detail below.

While ‘learning’ is something that (hopefully) happens in the classroom, the term has multiple realisations in the brain. To illustrate, here is a thumbnail sketch that identifies eight interacting learning systems in the brain: (a) There is a system for memorising specific moments, which produces episodic or autobiographical memory. This is the hippocampus and the structures around it. This system can change its connections very quickly to record snapshots; (b) The brain learns associations between perceptual information and motor responses. It spots complex spatial and temporal patterns within this knowledge, so-called ‘concepts’. This happens within the cortex, where changing connections takes seconds, minutes and hours; (c) Some associations are unconscious and involve the emotion (limbic) structures further inside the brain, associations between stimulus and response usually referred to as ‘classical conditioning’. These associations can form over seconds and minutes; (d) The brain learns to control content-specific systems in posterior cortex so that they are activated in the appropriate contexts. Control involves the prefrontal cortex, which also interacts with limbic structures to integrate planning with emotion; (e) There is a reward-based system that works out what we have to do to get what we want, to make nice things happen and avoid bad things happening, which operates over seconds and minutes; (f) There is a procedural learning system for learning activities that we perform frequently and often unconsciously, such as tying shoelaces, reading or driving a car. These automatic skills can take tens or hundreds of hours to learn through practice. The structures involved are the looping outer-to-inner circuits connecting the cortex through the basal ganglia to the thalamus and back again, and the cerebellum; (g) The brain can take advantage of its widespread circuits for perceiving and understanding other people, so that skills can be learned simply by observing other people, so-called ‘modelling’; (h) The brain can take advantage of its widespread circuits for using language to construct new concepts and plans, so that skills can be learned through instruction.

In addition to these multiple systems, a broader principle operates: to make all processes automatic, so they occur quickly, smoothly and without need for cognitive effort or even awareness. Skills are progressively transferred to basal ganglia and cerebellar structures. The more knowledge/skills are used, the more they become automatic. By the same token, the less they are used, the more the skill or knowledge is likely to be lost. Forgetting happens at a different pace in different learning systems. All of these systems work in an integrated fashion; they respond differently over time and to regimes of training; and they can be differentially modulated by other factors, such as motivational and emotional state. In face of this complexity, understanding the implications of this constellation of mechanisms for the term ‘learning’ as construed by educators is a huge challenge.

Even if educational neuroscience succeeds in this challenge, learning is only one part of education. Inspired by Bronfenbrenner's ecological systems theory (Bronfenbrenner, 1992), Figure 2 places learning outcomes at the heart of education, but illustrates the range of other factors, governmental, societal, institutional and child-internal, which make up the broader picture. As Bronfenbrenner's writings demonstrate, the factors that influence a child's learning outcomes, which operate at vastly different degrees of proximity to the learning process, should be seen as an interactive, interconnected system. The goal of educational neuroscience is to improve educational outcomes, largely by changing the most proximal factors to learning outcomes shown Figure 2: ability, motivation and attention, health and nutrition. However, it should be borne in mind the range of barriers to change that may be encountered beyond optimising learning itself; behavioural change needs to be considered within the wider framework of implementation science (e.g. Michie, van Stralen, & West, 2011).

Policymakers seek to put in place the right structures to deliver the educational outcomes desired by society. This involves deciding on types of schools, types of educators, subjects; when children should begin and leave school; who needs what qualifications and who needs what resources; the types of examination and assessment; and so forth. These decisions are predicated on having a clear goal for the education system. With clear goals, a study of learning mechanisms can inform the best way to reach them. However, the goals are not always clear. In particular, it is not clear whether policy changes are intended to improve outcomes for all, moving up the performance level of the whole population – for example, raising the mean performance of one country above another in international league tables – or to alter the gaps between children – for example, a policy to ‘leave no child behind’. Factors that influence population means may differ from those that alter the shape of population distributions, as research into historical changes in intelligence levels has illustrated (e.g. Flynn, 2012). For example, in industrialised nations, access to technology and improved nutrition have both been hypothesised as potential driving factors for population-level increases in IQ-test performance, but while experience with technology may act to move everybody's performance up, improvements in nutrition likely work to decrease variability in the distribution, preferentially benefitting those from lower SES backgrounds (see Pietschnig & Voracek, 2015).

We know that from multiple different perspectives, increasing mean education level translates to positive outcomes for societies (Lindahl & Krueger, 2001; Lutz, Creso Cuaresma, & Sanderson, 2008; Viner et al., 2017). However, the notion of mean education level conflates educational excellence with educational equality; high mean performance could be achieved either by the work of a few excellent performers or by low educational inequality with few poor performers (Schmidt & Burroughs, 2016). Hence, which educational strategies most effectively and efficiently promote growth? Is it wiser for countries to maximise performance at the top of the distribution at the risk of increasing the dispersion of performance or to minimise variability by boosting performance at the bottom? One international study of mathematics achievement in 13–14 year olds found a negative correlation between the dispersion of test scores and median test score for that country (Freeman, Machin, & Viarengo, 2010). Similarly, the mean educational achievement of a country has been found to be positively associated with a smaller impact of family background on achievement (Hermann & Horn, 2011). These studies suggest that countries gain more in terms of educational outcome by decreasing educational inequality, and that the study of learning mechanisms should inform policymakers as regards the most efficient ways to promote educational equality.

Finally, even with a solid understanding of the science of learning, translation into classroom practice is difficult. With its much longer, 125-year history of the study of learning, psychology still struggles to properly inform teaching practices. Techniques persist in the classroom even when there is a large body of evidence indicating a lack of effect (Roediger, 2013). For example, teachers still encourage students to highlight/underline text and reread text to enhance learning, despite evidence from psychology that neither is effective (Dunlosky, Rawson, Marsh, Nathan, & Willingham, 2013). It is not straightforward to translate an understanding of how learning occurs in the brain into ways to improve learning outcomes through different types of instruction nor is it straightforward to translate an understanding of the cause of learning impairments into appropriate interventions (Bowers, 2016a).

Part of the issue is to determine what teachers need to know about neuroscience or psychology theory. Teaching in the classroom is an interactive skill delivered in the moment, so it cannot be directly driven by theory (Howard-Jones et al., in press). Indeed, Willingham (2018) has recently argued that teachers do not even need to know psychological theory, let alone neuroscience. Rather, teachers should be familiar with behavioural observations – developmental patterns and consistencies in children's cognition, motivation and emotion, equating to “understanding children”. In this view, training for teachers should comprise practice in applying principles in context. Therefore, while it is intuitive that scientific insights into learning mechanisms should help improve educational outcomes, translational sciences have many hurdles to clear.