What tools does cognitive neuroscience use?

One of the most-cited applications of neuroscience to educational research has been that it contributes to our understanding of atypical development (Butterworth, Varma, & Laurillard, 2011; Gabrieli, 2009; Goswami, 2004; Royal Society, 2011). More specifically, disorders in the acquisition of school-relevant skills, such as dyslexia and dyscalculia, have been grouped under the term neurodevelopmental disorders, a category that also includes autism spectrum disorder, ADHD, and intellectual disability, all of which have serious consequences when the child enters school and for which educational interventions have been developed. The neuro in neurodevelopmental disorders refers to the idea that the origin of these disorders is biological and that they result from aberrant brain structure and/or function. The precise biological causes of these disorders remain unknown, but neuroimaging research has made considerable progress in understanding brain structure and function in these conditions. For example, research on dyslexia, a neurodevelopmental disorder that is characterized by specific and persistent deficits in learning to read, has revealed abnormalities in the brain networks that are used for the processing of phonemes (Eden, Olulade, Evans, Krafnick, & Alkire, 2016; Gabrieli, 2009), which is an important prerequisite for learning to read (Melby-Lervag, Lyster, & Hulme, 2012). Similarly, it has been shown that the brain networks that support the processing of numerical magnitudes, a skill that is very important for learning to calculate (De Smedt, Noël, Gilmore, & Ansari, 2013; Schneider et al., 2017), are impaired in individuals with dyscalculia, who experience serious and life-long difficulties in basic calculations (Butterworth et al., 2011). These neuroimaging data then also suggested that the cognitive skills that are subserved by these brain networks should be the particular focus of educational remedial interventions (Butterworth et al., 2011; Gabrieli, 2009; McCandliss, 2010; Shaywitz, Morris, & Shaywitz, 2009). Effective interventions that target specific deficits in phonological processing (e.g., Snowling & Hulme, 2011) and numerical magnitude understanding (e.g., Dyson, Jordan, & Glutting, 2013) in children with learning disabilities have been developed, and even their effects on brain structure and function have been studied (e.g., Barquero, Davis, & Cutting, 2014; Fraga González et al., 2016; Kucian et al., 2011).

On the other hand, it needs to be empirically verified whether the abnormalities in these brain networks are truly the cause of these learning disabilities or whether these brain abnormalities are the consequence of poor academic achievement. This remains unknown, as most of the existing body of data are correlational as well as cross-sectional, and they simply report associations between a particular disorder and brain abnormalities, which do not allow one to determine the direction of associations and their causality (Goswami, 2008). Interestingly, recent studies are now beginning to show that the brain abnormalities in phonological processing areas in dyslexia are already present before children learn to read, and that they predict later reading acquisition. These studies depart from the genetic nature of dyslexia (Snowling & Melby-Lervag, 2016) and compare children with a family risk for dyslexia (i.e., first-degree relative with dyslexia) to those without such a risk. Children with a family risk for dyslexia can be identified before the onset of formal education (e.g., in preschool), and their brain development can be characterized. As soon as it is possible to clinically diagnose dyslexia (i.e., in second grade), data can retrospectively be analyzed by comparing children with and without dyslexia before and during the early years of schooling. This allows one to disentangle causes (abnormalities that are already present before children learn to read) from consequences (abnormalities that emerge after schooling started, which may be the result of less reading experience). Family risk studies in dyslexia suggest that these brain abnormalities are likely, at least in part, to be the origin of their reading difficulties (Vandermosten, Hoeft, & Norton, 2016). In all, if these neurobiological causes of atypical development can be identified and can be detected at an early age, they can be used as markers to predict atypical development and, even further, educational outcome.

Neuro-Prediction

Neuro-prediction (De Smedt & Grabner, 2015), or neuroprognosis, refers to the application that brain imaging measures can be used as biomarkers to predict educational outcomes (e.g., Black, Myers, & Hoeft, 2015; Hoeft et al., 2007) and in particular to early identify children at risk for learning difficulties (Diamond & Amso, 2008; Gabrieli, 2009; Goswami, 2008). More specifically, brain imaging data can be collected before children possess skills that are necessary for traditional behavioral assessment, such as language. This allows the identification of at-risk children before the start of formal education and opens opportunities for early intervention, which may have preventive effects. Attempts to discover such biomarkers have been made in the early detection of dyslexia. Molfese (2000) showed that ERP responses to speech sounds recorded in newborns discriminated with 81% accuracy those infants who would develop dyslexia at the age of 8. Even though this classification accuracy is significantly beyond chance, it should be interpreted with great caution as it indicates that this biomarker incorrectly diagnoses approximately 20% of the cases (i.e., 20% false positives: incorrectly identifying children without dyslexia as having dyslexia; as well as 20% false negatives: failing to identify children with dyslexia as having dyslexia).

More recently, studies have started to investigate whether brain imaging measures can predict subsequent learning gains (Hoeft et al., 2011) or whether they even can predict response to educational interventions (Supekar et al., 2013). Supekar et al. (2013) investigated which brain measures, in addition to behavioral outcomes, predicted the gains of a one-on-one math tutoring intervention. Their data revealed that only volume and connectivity of the hippocampus, and not the behavioral data, predicted the learning gains: the larger the hippocampus before the start of the intervention, the larger the learning gains. This association of learning gains with the size of the hippocampus is not unsurprising, given that this area of the brain is particularly relevant to the consolidation of facts in memory, and that the specific educational intervention under study involved the automatization of arithmetic facts.

These are just the very early steps in trying to predict outcomes of educational interventions on the basis of neuroscientific data. The success of this approach will stand or fall on the quality of the educational interventions that are being investigated. This requires the involvement of educational researchers in these kinds of cognitive neuroscientific studies. Without this, there is a serious risk that such predictive studies are meaningless to both cognitive neuroscience and educational research, due to the lack of theoretical grounding of such studies. And even though brain imaging measures can predict later (reading) achievement, it will be important to determine the value added of these—cost intensive—measures on top of traditional behavioral assessments. There are preliminary data that indeed suggest that neuroimaging data can explain additional variance in academic achievement beyond what is predicted from behavioral measures (Hoeft et al., 2007), but this again depends on the careful selection of behavioral measures to predict a given behavior.

Learning at the Biological Level

Neuroimaging studies also allow us to understand learning at the biological level, which adds a new level of analysis to educational theory, for example in models on the acquisition of school-taught skills, such as reading and mathematics. This has the potential to complement as well as extend the existing knowledge that has been obtained on the basis of psychological educational research, and this new level of analysis might lead to a more complete understanding of learning (see also Lieberman, Schreiber, & Ochsner, 2003, for an application in political science).

One example is the componential understanding of complex cognitive skills taught via education (Dowker, 2005). Specifically, Dowker (2005) argued, against the background of cognitive neuropsychological studies with brain damaged patients (e.g., Dehaene & Cohen, 1997), that mathematics is not a unitary skill but instead should be conceived of as consisting of multiple components to which interventions should be tailored. Specifically, these neuropsychological studies showed that brain damaged patients can be selectively impaired in different yet specific areas of mathematics. This fractionation of mathematical skill has then be used to investigate individual differences in these components of mathematical skill and to develop educational interventions that take these components as starting points to tailor interventions to the specific strengths and weaknesses of children in mathematics (Dowker, 2005).

Neuroimaging data can also provide construct validity of educational theories. For example, the IQ-discrepancy criterion—the observation that in children with specific learning disorders there should be a clear discrepancy between their IQ and their academic performance—has been criticized for many years in psychological and educational research. This critique has been validated in behavioral studies by showing that individuals with and without a discrepancy between their IQ and poor academic performance do not differ in their cognitive profile and in their response to educational interventions (Fletcher, Lyon, Fuchs, & Barnes, 2007). Tanaka et al. (2011) tested the validity of this observation at the biological level. They compared the brain activity during reading in children with reading difficulties with and without a discrepancy with their IQ. Their findings revealed that the brain activity patterns in the two groups of children did not differ, which converges with the behavioral data that reading difficulties (and their neural correlates) are independent of IQ. These observations also validated the removal of the IQ-discrepancy criterion in the definition of specific learning disorders in the latest version of DSM-V (Diagnostic and Statistical Manual, American Psychiatric Association, 2013), a classification of mental disorders widely used by mental health providers.

This observation of converging evidence has been belittled or disparaged by critics of educational neuroscience (e.g., Bowers, 2016; Smeyers, 2016), who argue that such findings add nothing to what we already know on the basis of psychological or educational research. However, if a particular mental process has an identified biological substrate, then the theoretical understanding of this process will have more exploratory power if it is constrained by both behavioral and biological data; and consequently, a better explanatory model for a given educational phenomenon will be a better base for grounding educational interventions (Howard-Jones et al., 2016).

There are also examples of studies in which neuroimaging data reveal something different (diverging evidence) from what is observed from behavioral data. This is particularly the case when one aims to measure subtle processes that are hard to capture with behavioral data alone. Investigating optimal ways to foster foreign language learning, Morgan-Short, Steinhauer, Sanz, and Ullman (2012) used ERPs to unravel the differential effects of explicit language education (grammar-focused classroom setting) versus implicit education (immersion setting) on syntactic processing. While both approaches resulted in similar behavioral improvements, as measured by means of a cognitive task in which participants had to judge the correctness of sentences, ERP measures revealed striking differences between the two approaches. The immersion setting resulted in a brain response that was similar to what is observed in native speakers, whereas this was not the case in the grammar-focused setting, suggesting that the first approach may be more beneficial in adult foreign-language learning than the latter.

More recently, Anderson, Pyke, and Fincham (2016) demonstrated by using multivariate fMRI analyses that it is possible to parse the online problem-solving process of a particular mathematical problem into different cognitive stages (i.e., encoding, planning, solving, and responding). This offers an interesting possibility to investigate the durations of each of these stages and how they change as a function of (different) educational approaches. These processes are very difficult, if not impossible, to capture by means of behavioral data, such as systematic error analysis, the analysis of latencies, or the use of introspective verbal protocols.

Another example of neuroimaging methods that reveal something different than can be observed from behavioral data lies in the detection of compensatory processes that arise in the context of remedial interventions. Various intervention studies in the domain of reading have revealed that these evidence-based programs lead to a normalization of brain activity and structure in those networks that are typically associated with normal reading, and this is accompanied by improvements in behavioral reading performance (e.g., Hoeft et al., 2011; Keller & Just, 2009; Temple et al., 2003). On the other hand, each of these studies also revealed changes in brain circuits not typically associated with reading, such as changes in the right prefrontal cortex, an observation that is suggestive of the involvement of compensatory processes. Such compensatory mechanisms are hard to detect via behavioral data. The precise function of such compensatory processes is often unclear and needs to be better understood but offers a promising avenue for informing future interventions.

Generating Predictions for Educational Research

This study of compensatory processes is also an example of a promising application of cognitive neuroscience to educational research that is more indirect. Specifically, findings from cognitive neuroscience might have the potential to generate new hypotheses on educational phenomena that can be tested in follow-up educational research, and a similar application of neuroscience to psychological research has been described by Aue, Lavelle, and Cacioppo (2009). In this way, an iterative cycle of interdisciplinary research can be generated (see also Howard-Jones et al., 2016). One example comes from the study of the role of finger representations in numerical development (Kaufmann et al., 2008). These authors examined brain activity during the comparison of numbers, a crucial skill in mathematical development (e.g., De Smedt et al., 2013), in children and adults. While the groups did not differ in their behavioral performance, children showed more activation in those brain areas that are associated with finger movements and grasping, leading Kaufmann et al. (2008) to suggest that finger-based representations play a more important role in children’s understanding of number, and perhaps should be addressed when designing interventions. The role of finger representations in numerical development has been the focus of a series of recent behavioral studies, but the evidence is mixed to date (Long et al., 2016; Wasner et al., 2016), leaving it unresolved whether the use of fingers should be encouraged or discouraged when teaching early mathematics.

Effects of Education on Biology

One of the key findings in neuroscience is that our brains are highly plastic, which means that they are shaped by experience, a process referred to as experience-dependent plasticity (Diamond & Amso, 2008; Johnson & De Haan, 2011), and this process is present throughout life. There are massive developmental changes in brain structure and function that continue into late adolescence (Giedd & Rapoport, 2010) and beyond, and these changes are driven by environmental input. One example comes from a study of London taxi drivers (Maguire et al., 2000) in which the effect of extensive training in learning how to navigate in the city of London on brain structure was investigated. This study revealed that training navigational skills induced changes in the hippocampi, a structure that is also involved in spatial navigation, of these taxi drivers and that the amount of training correlated with the size of the observed morphological changes in the brain. Importantly, there were individual differences in the extent to which the training affected brain structure, which suggests that plasticity is not unlimited.

Because children spend a large amount of time at school, education is one of the most powerful sources that shapes the development of our brains. There is now an increasing number of studies that investigate how learning to read changes and reorganizes brain structure and function (Dehaene, Cohen, Morais, & Kolinsky, 2015; Skeide et al., 2017). Literacy acquisition not only constructs brain circuits that become associated with reading but also changes brain circuitry (as well as its connectivity) that is not typically associated with reading, such as the visual ventral stream. These changes have also been observed in formerly illiterate adults, which exemplifies that plastic brain changes can occur at different ages (Dehaene, Cohen, Morais, & Kolinsky, 2015; Skeide et al., 2017). Another example is provided by Neville et al. (2013), who designed a family-based intervention program, based on knowledge of the neuroplasticity of attention and parenting research. They showed positive effects of the intervention on preschoolers from low socioeconomic backgrounds, and these effects were observed in electrophysiological measures of brain function, cognitive measures, and parent reports on child behavior. What is currently missing are studies that investigate what precise aspects of these educational experiences affect brain development and the limits of this plasticity, and this clearly represents an area for future studies at the crossroads of cognitive neuroscience and educational research.

Special education represents another area in which the effects of education on the brain have been revealed, through the study of specific remedial interventions on brain structure and function in atypical development (e.g., McCandliss, 2010). Such studies have revealed that processes of normalization—brain function becomes more similar to a typically developing control group—and compensation—activity patterns in regions different from what is observed in typically developing children—occur. These patterns on how individuals with atypical learning compensate for their difficulties are very relevant to education, as they may provide novel ways of teaching specific compensation strategies, the effects of which should be investigated in educational research.

It is important to emphasize that studies that investigate the effects of education on the brain should carefully take into account the broader educational context (e.g., participants’ learning histories, teaching materials) in which learning takes place. These variables should not be considered as confounds that should be controlled for. Rather, they should be the focus of interest as variability in these factors will have massive effects on brain structure and function. Future studies should therefore consider how these characteristics of the learning context moderate neural data acquired via brain imaging measures.

Biological Interventions

In addition to investigating the effects of educational interventions on brain structure and function, recent advances have made it possible to directly intervene at the biological level and to use neurophysiological interventions, that is, transcranial electrical brain stimulation or TES, to directly affect brain activity and consequently affect behavior through the change of brain function (Cohen-Kadosh, 2014). During TES, a small electrical current is non-invasively applied to the brain via electrodes fixated at the scalp. The current is thought to change the activity level of the cortical regions that are under the electrodes and is assumed to change performance or learning. Various studies show the effects of TES (Cohen-Kadosh, 2014, for a review). For example, the use of particular types of brain stimulation leads to improved performance in arithmetic (Krause & Cohen-Kadosh, 2013), although not all individuals respond in the same way to such stimulation (Krause & Cohen-Kadosh, 2014).

The field of brain stimulation is currently in its infancy, and at this point we do not fully understand the underlying mechanisms of these techniques (Schuijer, de Jong, Kupper, & van Atteveldt, 2017). It is also crucial to emphasize that TES has an effect only if it is accompanied by traditional behavioral and cognitive training. The existing studies are limited to (mainly healthy) adults and are not applicable to the developing brain, and it remains to be verified if such education-related applications are ethically possible (Schuijer et al., 2017).

What are tools of cognitive neuroscience?

Which technology is used in cognitive neuroscience to examine structures in the brain?

What is cognitive neuroscience approach?

What is an example of cognitive neuroscience?