Educators, politicians, students and parents have diverse ideas about the ways in which individuals learn and can be taught, but these ideas are not always backed up by empirical research. Educational psychology has its roots in educational theory rather than experimental psychology, and is one area of expertise that can provide evidence to inform educational practice. In this book we consider some of the psychological functions that are particularly important for education – language acquisition, learning and memory, ability, sex differences and creativity. In each case, we present an account of the basic psychology (and where necessary, neurology1) related to this area, alongside seminal studies and cutting-edge research that link the psychological knowledge back to education. In later chapters, we look at particular applied areas in the field of educational research: reading, the role of metacognition (thinking about learning), the effects of academic selection, the changes in cognition that occur with ageing, and the role of technology in the classroom. We consider how these areas are treated in different countries across the world and how they impact on social policy in England.2

We have, of course, been highly selective in choosing which studies to report, only including those which we feel best illustrate the points we are making. In some cases considerable detail is provided, in order to give the reader a chance to evaluate the evidence presented and form an independent opinion on its implications. References to the relevant scholarly research papers are also included in order to make it easier for academic readers to follow up on our conclusions. A summary at the end of each chapter addresses the question so often asked when new research findings are presented: ‘So what?’ We review the ways in which we believe the psychology we have described could support, and in some cases possibly change, the practice of teaching.

This book is intended not only for teachers and those studying education as an academic subject, but also for parents, grandparents and others who are interested in the education of the children and young people of today.

What Do We Mean by ‘Evidence’?

This section is intended for any of our readers who are not familiar with reading articles in scientific journals.

Throughout this book we have emphasized the importance of evidence for the effect of interventions and for demonstrating the detection of differences in education. We do, however, need to make clear what we are looking for when we present the findings of a scientific paper. We have sought to present empirical evidence which relates to contemporary methods of teaching and to the organization of our education system. However, it is essential to view such evidence critically, because at times researchers are just as likely as politicians to present data in a way that serves their own interests rather than giving a balanced view. So what sort of things do you need to look out for when you read about studies in this book and elsewhere?

Firstly, there are some basic aspects of research design of which you ought to be aware. Unless they use census data, all studies involve a sample of people and it is important to consider the nature of this sample if we want to know whether the results apply to the general population. How were participants recruited? By advertising, for example, or by buttonholing people in a supermarket, by using all the available children in a school year, or by following up particular birth cohorts? Actually, quite a lot of psychological research is done using university students. A second question to ask is what selection criteria were used and what proportion of the original sample was discarded as a result of applying these criteria. For example, in the study of ageing, were elderly people with organic disease or dementia included or not? Different methods of obtaining and selecting participants will introduce different possible biases.

Research can range from the highly qualitative to highly quantitative, with every mixture imaginable in between. Such studies can tell you different things. Qualitative studies, for example case studies or a small number of in-depth interviews, are very helpful for exploring people's motivations, why they do things, but are not so good for identifying general rules. We cannot necessarily assume that one set of people will behave like another. In contrast, large quantitative surveys can be good for capturing representative views and for identifying patterns of behaviour but they rarely tell us why these patterns occur. Large sample sizes are good because then small individual differences or errors will have less of an impact on the results. However, they also present a risk because with large samples many differences can be statistically significant without necessarily being that important.

When doing quantitative research most researchers will run statistical tests such as t-tests or ANOVAs (which tell us whether differences between average group results are significant) and correlations or regressions (which tell us if two factors vary together). These typically produce a ‘p’ value between 0 and 1. A p value of 0.05 indicates there is a 1 in 20 probability that the result occurred by chance and is a common criterion for statistical significance. It is worth noting, therefore, that if a researcher simply runs endless statistical tests then eventually they are likely to get a significant result by chance. In other words, don't assume that a significant result proves something, instead try to think about what it means and whether it actually makes sense.

Another caution is in the temptation to over-state findings, particularly when talking about causality. The key mantra is that correlation does not equal causation. In other words, although two things might vary together, it does not prove that one causes the other. You might think this is obvious, but it is easy to be convinced by an argument that intuitively sounds correct. For example, it is easy to note that socioeconomic status correlates with many educational outcomes such as attainment and the likelihood of going to university. However, being of low socioeconomic status does not in itself cause a pupil to have lower attainment. It is the various associated factors, for example the likelihood of parents reading to their children or the likelihood that a parent can pay for extra tuition, which actually have a direct impact on attainment.

Studies can either be cross-sectional – a snapshot in time – or longitudinal, with measures taken before and after a time interval. If we want to demonstrate causation then either we can use qualitative interviews to ask why people do things (and rely on the accuracy of introspection) or we have to conduct a formal, longitudinal experiment where we try to hold as many things constant as possible and then vary the factor of interest. If we start with two similar groups and find that after an intervention with one of them they differ significantly, there is a good chance this can be attributed to the intervention. However, it is important to be sure that the two groups were matched on relevant factors. In education such factors that are likely to affect outcome are: measured ‘ability’; socioeconomic background; parents' educational level; age; and sex. Of course, whilst this approach is ideal when working with levers, cells or chemicals, it is impossible in practice to find two identical classes and teachers, so these experiments are always open to critique. One should also check whether any control group undertook a comparable activity to the intervention activity, but one not targeted at the outcome of interest. This accounts for ‘tender loving care’ effects – that people can change their behaviour just because they know they are being studied. Longitudinal studies make the strongest case for proving causation but they are never perfect. This is one reason why it is so important to conduct multiple studies and replicate findings.

Unfortunately, it is very difficult for academics to get studies published if all they do is replicate the work of someone else. For this reason, we tend to get a lot of research that is similar but not identical. If we still want to pick out patterns across the papers, then one technique available is meta-analysis. Here, a researcher will collect together multiple papers addressing a single topic and try to aggregate the findings to see whether overall they are positive, negative or lack a clear pattern. This method has a lot of potential, but it relies on the researcher collecting a comprehensive sample of literature, filtering out poor quality studies and weighting the remainder to account for factors like sample size. It is also susceptible to problems of publication bias, because it is much less likely that a study will get published, and therefore included in the meta-analysis, if it includes no significant findings (after all, would you be more interested in reading about a food that boosts attention span or one that has absolutely no impact on it?). A good meta-analysis provides a helpful aggregation of literature, but you should not take the findings as fact.

This, actually, is the core message when reading empirical research: just because a researcher writes a very convincing paper highlighting the importance and relevance of their findings, you should not assume they have proved anything. You must always think through the logic of the study and consider every possible alternative explanation before coming to your own conclusion about what it means.