Insights from Learning as a Generative Activity Part 1: What is Learning?

Over the last few years, countless blogs, articles and books have been written, each one attempting to distill and describe the essence of good teaching. There has been a major shift in people’s opinions as explicit instruction has been lifted out of the shadows and placed at the centre of many school’s pedagogical frameworks. This shift is partly down to the massively increased interest in Rosenshine’s Principles of Instruction, a framework that has become almost ubiquitous now, partly thanks to the work of Tom Sherrington and Oliver Caviglioli.

Rosenshine’s work is based upon three broad areas: research into cognitive science; research on the classroom practices of expert teachers and also research on cognitive supports like scaffolding and using models to help students learn complex tasks. This AFT article from Spring 2012 is a good starting point if you want an overview of Rosenshine’s ideas and research.


While Rosenshine’s work can be seen as research-informed overview of the kind of approaches and techniques that effective teachers use, Learning as a Generative Activity outlines eight different ways to foster ‘generative learning’ in students. Much like Strengthening the Student Toolbox, Learning as a Generative Activity explores strategies that students can use in order to learn effectively. Mayer’s work should be seen as complementary to Rosenshine’s as the instructional choices of teachers will determine the approaches that students adopt. Effective instructional choices will be based upon a sound understanding of how students learn and effective teachers will direct students to interact with content in ways that help students ‘make sense of the material’.

What is Learning?

The authors of Learning as a Generative Activity, Mayer and Fiorelli, define ‘generative learning’ as ‘helping learners to actively make sense of the material so they can build meaningful learning activities to transfer what they have learned to solving new problems.’ This definition is far more detailed than Kirschner’s proposition that ‘learning is a change in long term memory’ and the additional requirement of transfer means that many would reject it  as being too exacting. As Dylan Wiliam has pointed out, if transfer is a necessary condition of learning, then this would imply that memorising your social security number is not an example of learning.

Transfer is notoriously hard to achieve, even within the same domain and at best, we should be attempting to induce near transfer in our students so that their understanding moves from being inflexible and shallow to flexible and deep. David Didau gives a useful overview of the concept of transfer in this blog.

The tracks within DI schemes are planned in order to promote this increase in flexibility and the purpose of teaching to the general case-one of the main aims of DI-is to ensure that students can apply their knowledge to new but similar problems or applications. This can be achieved by manipulating sequences of examples or by making students aware of the similarities and differences between different contexts-essentially helping them to perceive the deep structure of a problem. A well planned instructional sequence will purposively facilitate this transition from inflexible to flexible knowledge as students move from restricted drills to freer application. The table at the end of this post demonstrates this transition. Initial tasks may look much like rote memorization where students merely respond to a cue-perhaps asking student to recall a specific date or fact; later tasks may ask students to develop their understanding-perhaps using elaborative interrogation; later still, tasks may require students to apply these component parts within a wider application-perhaps using the date or fact as part of a chain historical argument. 

So if transfer or flexible understanding are important aspects of learning, why has Kirschner’s definition become so widely accepted? I would argue that his definition has two main strengths. Firstly, its succinct nature encourages teachers to concentrate on the vital role of memory and retention in learning and, in an age where attainment is often judged via close books exams, this can be no bad thing. Secondly, its breadth allows it to encompass the entire continuum from inflexible to flexible knowledge. Both ends of this continuum have important roles to play in learning, and, while ‘flexible knowledge’ is the end goal of almost all instructional sequences, ‘inflexible knowledge’ almost always forms the foundations of understanding. If a definition of learning effectively excludes ‘inflexible knowledge’ by adding the additional requirement of transfer, then this may have adverse effects: teachers may come to believe that rote learning is undesirable or ineffective; equally, teachers might focus entirely on novel problem solving in the belief that the best way to improve this ability is to practice problem solving itself when, counterintuitively, it may be more effective to begin by building student background knowledge which will initially be inflexible. Under Kirschner’s definition, both the rote memorization of your social security number and the ability to solve novel algebraic problems would be defined as learning.

So if Kirschner’s definition has clear strengths, what else can we learn from Fiorella and Mayer’s? Their definition of ‘generative learning’ is ‘helping learners to actively make sense of the material so they can build meaningful learning activities to transfer what they have learned to solving new problems.’ This definition of learning seems aimed at the end goal of most instructional sequences: flexible knowledge and understanding in which students can achieve near transfer. This seems entirely uncontroversial: we want students to be able to flexibly apply their knowledge. When we first learn a concept, our understanding is shallow, especially if it is a concept unrelated to our existing background knowledge. As we begin to make sense of what we have learned, we make connections between the newly learnt concept and our existing knowledge.

In an attempt to further hone their conception of learning, the authors critique a number of others:

Conception 1: Learning works by engaging in hands-on activity, so it is better for you to learn by doing rather than by being told.

Mayer and Fiorelli dismiss this as being overly concerned with behavioural activity and the nature of the learning tasks that student attempt. They point out that this conception fails to focus ‘enough on cognitive activity’ as students could be busily engaged in an activity without purposively thinking about the content in a way that helps them make sense of it. This opinion is analogous to Rob Coe’s assertion that ‘learning happens when you think hard.’ When teachers focus on the activity rather than what students are thinking about, this often results in classrooms that seem busy and purposive but lack any real cognitive processing. A good piece of advice when planning lessons is to always think about what the students will be thinking about. Activities or tasks should be chosen so students can spend as much time as possible thinking as deeply as possible about the content. If seen through the lens of Cognitive Load Theory, tasks should be easy to explain to students with no irrelevant procedures or activities that can either cause confusion or distract students from thinking about the content that they need to ‘actively make sense of.’

Conception 2: Learning works by building association, so you should practice giving the right response over and over.

The authors dismiss this conception of learning as being too prohibitive and only really applying to situations where students are expected to ‘give the right response for a given stimulus’. They concede that this conception is not wrong, but that it is ‘just too limited’ as it in no way ‘deals with understanding’ which they define as when people are able to ‘take what they have learned and apply it in new situations.’ While there are many instances of learning that are described by associative learning (Dylan Wiliam’s social security example fits this well), if this conception were to be a catch all definition, it would fail to describe what is the end goal of most instructional sequences: generalized, flexible understanding.

As an aside, the very term ‘understanding’ seems to be contested: when Fiorellia and Mayer use it, they intend it to include near transfer; Willingham, however, gives a slightly more restrictive definition, defining it as ‘remembering in disguise…What do cognitive psychologists know about how students understand things? The answer is that they understand new ideas (things they don’t know) by relating them to old ideas (things they do know)’ Interestingly, this definition fits exactly with the three stages of cognitive processing that form the theoretical basis of ‘sense making’ and underpin all of the learning strategies that Learning as A Generative Activity explores:

  1. Selecting relevant information to attend to
  2. Organising the material into a coherent cognitive structure in working memory
  3. Integrating it with relevant prior knowledge activated from long term memory

Conception 3: Learning works by adding information to your memory, so you should work hard to find and memorise new information

The authors dismiss this conception because ‘humans do not work like computers’ and ‘we do not simply take in what was presented and put it in our memory’. They also posit that learning involves ‘changing what is presented from information (which is objective) into knowledge (which is subjective). I would argue that it is precisely because we are not ‘like computers’ that we need to adopt proven strategies like distributed practice and the testing effect in order to aid us with encoding and retaining information in our long term memories. While a computer reliably stores information without error as soon as it is asked to, our messy and fallible cognitive processes require the adoption of deliberate and specific learning strategies if we are to retain information. Cognitive Load Theory highlights the absolute importance of background knowledge, making the important point that ‘novices use thinking skills, experts use knowledge’ when attempting problems. Background knowledge also has a huge role in reading comprehension. Memorising or automatizing information can give a huge advantage to students. Times tables are a good example here: even simple algebraic problems are made impossible if students do not have reliable and rapid recall of multiplication facts. No-one, however, would argue that memorisation encompasses learning as a whole.

Conception 4: Learning is a social activity, so it is better for you to learn with others in a group than to learn alone.

Although the authors concede that ‘you can interact with others during the learning process’, and that social activities that promote meaningful learning can be effective, they dismiss this conception because group work is hard to manage and ‘research on group learning tends to show that all group interactions are not equally helpful’. Although group work can be useful if there are both group goals and individual accountability, it can result in social loafing and off-task behaviour. I tend to avoid group work for these very reasons although this does not mean that group work is a poor instructional strategy per se.

In the next post, I will continue to explore Fiorella and Mayer’s conception of learning.