Annual Research Review: Educational neuroscience: progress and prospects

Educational neuroscience has received diverse critiques, challenges and sources of support. Some of these do not pertain to the field of science itself. For example, there is a great deal of research into so-called neuromyths. Neuromyths are misconceptions about the brain and education among teachers and the public that are either not yet supported by the data or actively contradicted by existing science (e.g. Goswami, 2006; Howard-Jones, 2014a; Macdonald, Germine, Anderson, Christodoulou, & McGrath, 2017); examples include large benefits of students adopting a ‘growth mindset’ (Sisk, Burgoyne, Sun, Butler, & Macnamara, 2018) or teachers matching the presentation of teaching materials to students' individual ‘learning styles’ (Rohrer & Pashler, 2012). While this is an important issue in the communication of science and for the interaction between the stakeholders within educational neuroscience, it is not core to the enterprise of understanding the actual neuroscience of education, nor are miscommunications of science unique to educational neuroscience. Then, there is research on the so-called seductive allure of neuroscience, the finding that proposed new teaching techniques are more likely to be believed when accompanied by brain images (Farah & Hook, 2013; Weisberg, Keil, Goodstein, Rawson, & Gray, 2008). While this is an important issue of contextual framing in education research, again it does not concern the enterprise of educational neuroscience itself.

The substantive criticisms of the field to date come in three types: a priori arguments against the relevance of neuroscience to education, criticisms of the current practical operation of the field and doubts about the viability of neuroscience methods for diagnosis of disorders or prediction of individual differences.

Regarding the first type of criticism, some commentators have viewed neuroscience data as in principle not relevant to education. For example, Bowers (2016a) views education as solely concerning the effect of instruction on behavioural outcomes and does not view the underpinning neural mechanisms of learning as pertinent to this question. “My claim is that neuroscience is irrelevant to the design and evaluation of teaching” (Bowers, 2016b, p. 630). As we saw with Willingham's (2018) recommendations, this crit icism may also be levelled at psychology. A variation of this criticism is that (perhaps thus far) neuroscience has not led to new methods or insights that were not already suspected based on behavioural evidence, or not in need of verification using behavioural evidence. The measurement and assessment of behaviour, in this view, are solely the purview of psychology. This criticism is sometimes exaggerated by equating neuroscience with functional brain imaging (rather than understanding the discipline to be the investigation of neural mechanisms by a diverse set of methods, see footnote 3); if education is about behavioural change, it surely does not matter whether or not there are correlated changes in brain activity (see, e.g. Bishop, 2014). A more philosophical variety of the criticism is that educational neuroscience is inappropriately reductionist: to the extent education is a social phenomenon, social problems require social solutions, not reduction to neural mechanisms (e.g. Breckler, 2006; Lalancette & Campbell, 2012; for discussion). These criticisms broadly represent a resistance to interdisciplinary research. In the earlier section, ‘Why psychology is not enough', we indicated possible disadvantages to this resistance.

Regarding the second type, there have been practical criticisms about the way research is conducted and the pace of progress (e.g. Anderson & Della Sala, 2012). Much of the initial research in the field is likely to focus on understanding why the teaching methods that work do work, but there are expectations that neuroscience must quickly yield revolutionary teaching methods of large effect size (Thomas, 2013). Additionally, the methods required to collect neuroscience data, such as brain imaging, require controlled experimental situations. These are far from the context of naturalistic behaviour in the classroom and therefore of questionable validity. Furthermore, the neuroscience data themselves can be complex and hard to interpret. In terms of interdisciplinary dynamics, most educational neuroscientists advocate a dialogue between disciplines, but critics suggest that so far, it has constituted neuroscientists (patronisingly) lecturing teachers about what they should be doing in the classroom, with insufficient driving of the neuroscience research agenda by teachers (e.g. Turner, 2011). Where neuroscience has influenced educational policy and practice, it has been premature (Bruer, 1997). Broadly, these criticisms represent the growing pains of the field and areas where it must improve.

The last type of criticism has been in response to the proposed use of neuroscience data to predict developmental outcomes, such as dyslexia or autism, for example, using functional brain imaging or genetics. Neuroscience methods offer some advantages: they can distinguish different underlying causes of the same behaviour (Sheridan & McLaughlin, 2016), they can reveal processes not evident in behaviour (e.g. inhibition of task-irrelevant responses), and when collected early, they can have predictive power for developmental outcomes. However, it is argued that neuroscience methods do not yet have the specificity or sensitivity to be useful screening methods, that their cost is too high, and that practicality much poorer compared to existing, simpler behavioural methods (e.g. Bishop, 2013, 2014). These are reasonable criticisms of the current state of the technology and evidence base but may not always be so. Neuroscientific measures have the potential to complement behavioural or environmental risk factors by being available long before a child starts school, for example, genetic tests at birth, or electrophysiological measures of language processing during infancy, may predict dyslexia risk (e.g. Guttorm, Leppänen, Hämäläinen, Eklund, & Lyytinen, 2009). Early predictions will maximise the available time for intervention and may be more effective by virtue of being implemented earlier in development, reducing risk when children start school.