Dead Poets and Trunchbulls: Preliminary thoughts on Recognition and Recall.

-->

In real-life we are sometimes required to recognise something, on other occasions to recall it. Probably we do much more of the former than the latter – for example recognising something that is bad to eat, the face of a friend, a path through the woods.

The 'recall' process we might use in describing the creature that attacked us, the face of a mugger, the route to take to get to work.

Not only are the two processes quite different, the latter is much harder and notoriously unreliable. My aim here is to explain why this is, and suggest some implications for the way we design learning.

Recognition first: as we experience the world, we encode those experiences as increasingly subtle affective responses. As an infant our reactions to things will be nearer their homeostatic roots – fouls tastes, loud noises etc. As an adult we will have learned to react differently to subtler stimuli - architectural styles for example: some people respond with delight to Art Deco.

These sets of reactions become associated with sounds. When we experience a given reaction we are able to make the corresponding sound: such as ‘chair’.

Recognition proceeds in this fashion – experiences trigger reactions and if they are sufficiently close to a stored reaction, the experience is recognised. Note that this is a ‘fuzzy logic’ implementation: an experience is not required to have any specific set of features, but merely to cause a sufficiently similar affective response – and of course that will vary from one individual to the next. It is also true that as stored reactions fade over time, we can fail to recognise something we once knew.

Recall is quite a different process. In the case of recall, a word may be used which corresponds to the stored reaction set. The affective responses are activated, though not as strongly as if the individual were actually having the experience. The affective response allows the individual to recreate the experience to some degree – for example activating similar areas of the visual cortex, creating something a bit like an ‘indistinct sketch’. Even for objects and experiences that are very familiar, this process is remarkably inaccurate – for example the words ‘your house’ may conjure up an image, but it is likely that you will get many features of your house wrong, despite being intimately familiar with it.

If you have a pet dog, something similar probably happens when you say the word ‘walk’ – the words triggers a set of emotional responses that allow your dog to reconstruct the experience you are describing – and the two of you to communicate. Note that it would be relatively easy to train your dog to correctly complete multiple-choice tests - but much harder to get it to write an essay. 

Recall can be improved by strengthening the affective significance of experiences – for example the ‘memory palace’ technique in which people create bizarre visual scenes along a familiar path or placed in a familiar location (e.g. the rooms in their home) in order to create strong affective cues. For example, if one imagines a monkey in the shower – which would surely be a shocking and emotional encounter – one can more easily recall this scene, and the cue ‘monkey’ which might then trigger the word ‘key’.

Despite this, recall is generally very weak, since it relies entirely on confabulation – i.e. the reconstruction of experiences from their affective impact. Errors in recall reflect the nature of this process: frightening creatures being recalled as bigger or more ferocious than they actually were, for example.

In conventional education, a central problem has been how to get individuals to recognise or recall factual information.

This problem is more complicated than it might first appear:

1. 1)    Firstly, people will often have very little reaction to much of the information to which they are exposed. In normal life, experiences to which we have little or no reaction are quickly forgotten or go entirely unnoticed. In education, however, people are expected to recall information that causes no response, because we have in many cases a ‘curriculum’ which doesn’t take into account the things that people care about. If they are especially passionate about medieval royalty for example, this is a happy accident.
2. 2)    Secondly, humans are not well adapted to learn in educational contexts or from educational content. Like other mammals, they are designed to move freely about their environment learning from experiences that impact their senses in a variety of immediate ways – smells, sights, sounds and feelings. In educational contexts people are often expected to learn while sitting in a single location, and from the appearance of words in a book alone.

That we can do this at all is truly extraordinary. How do we do it? As we read the words in a book we are able (sometimes with help) to confabulate scenes and experiences which cause some faint affective response (which may or may not resemble the real responses), and these reactions are used to encode the information. In this way we are able to learn about a volcano from a book, whilst sitting in a classroom. This affective responses tend to decay over time - faint ones very quickly - so repeating the activation at intervals will help somewhat. However, much of the research concerning spaced learning concerns information that would not normally be remembered and so only reveals how human memory works under abnormal conditions (which is a bit like studying fish respiration, when it is out of water).

One largely uncontrolled variable is, of course, the individual – and whether or not they are the kind of person that cares sufficiently about thing like volcanos to have a reaction to them.

Another significant variable is the approach which, in general terms, seeks to add affective significance to what is generally a very weak and uninteresting stimulus – such as a word on a page.

Broadly speaking, there are two approaches, which I shall call ‘Dead Poets’ and ‘Trunchbull’. Dead Poets is an approach which adds affective significance through a variety of mechanisms aimed at elaborating and deepening reactions to the material – for example, through storytelling, through heated debate, through sheer enthusiasm and the use of suggestive imagery – and, of course, by relating the words to deeper more instinctive concerns - such as love and heroism.

By contrast, the ‘Trunchbull’ approach depends largely on a range of anxieties to improve memory – such as anxiety regarding failure, tests, parental disappointment, future job prospects – and even more proximate anxieties such as bullying or actual or threatened violence.

Both mechanisms work in fundamentally similar ways: attaching affective significance to material which is otherwise insignificant – but are likely to have different results and associated behaviours. For example in the ‘Trunchbull’ condition, people are more likely to ask ‘will this be on the test?’ since their primary concern is whether or not information will be tested, and if it will, this generates sufficient anxiety to aid recall (much as a mammal will be more likely to remember a type of insect that stung them).

When it comes to recognition (for example in multiple-choice tests) the person undergoes the following process: each of the stimuli is scanned and the emotional response gauged. For example if the question is ‘Which of the following is the capital of France?’ and the responses ‘Berlin, Timbuctoo, Paris’ the person will read each response in turn and have a corresponding affective response in each case.

For students educated in the ‘Trunchbull’ style, a significant part of this affective response will be anxiety – the specific kind of ‘this will be on the test’ anxiety that the technique generates. On this basis we can make a couple of predictions: firstly, these kinds of students will perform worse when the other distractor responses are also items that are on the test than if they are not, since the ‘this is on the test’ anxiety is not a differentiator. Secondly, ‘Trunchbull’ students will be more likely to forget the information after (successfully) passing the test since this affective component used to store the information is now redundant.

Students educated on the ‘Dead Poets’ approach do not have the benefit of ‘stimulus anxiety’ to differentiate responses – instead each response will conjure up a richer affective response to each stimulus. This may not necessarily be a faster process, but it does predict that these individuals will be more likely to retain the information beyond the test.

Either way, multiple choice questions merely require students to match the affective response elicited by a stimulus (the responses) to a stored response. At a bare minimum this might merely be having a 'this is a test response'-type anxiety to one stimulus and not to the others. The difficulty with this educational approach is that it is ‘merely test-passing’: i.e. the response that someone has to a word describing an experience may be very different to the response a person has to the actual experience. A person may answer correctly that the appropriate lever to pull in an emergency is the ‘flight stabilisation lever’ but how they feel about the words ‘flight stabilisation lever’ and the actual lever itself (which they may never have actually experienced) can remain entirely distinct. Even worse, the affective responses to a word, on a piece of paper, in an exam room may be nearly entirely different to the affective responses a person experience in real life, where they are supposed to apply that knowledge. In this way people can give the impression that they are learning, whilst not actually learning anything – they are responding to words, not experiences, and how those words correspond to experiences will be highly variable. This is why it is so important that if we are to genuinely learn, rather than merely going through an educational ritual, that our learning experiences resemble the performance experience (affectively). Simulations are far better than multiple-choice questions for this reason,

In the case of recall, students educated on the ‘Trunchbull’ model will not have the advantage. Whilst the ‘Dead Poets’ approach is typically more labour-intensive, requiring extensive and varied affective elaboration, it permits the person to make more connections and to recall in greater detail – a person who is truly passionate about a topic will forge more connections than someone who is simply memorising information for a test. Once again, a significant differentiator is what happens to the new information post-test.

Recall is also a very different process to recognition: in recall, information must be reconstructed from the affective traces it left at presentation. This means that affective differentiation – the ‘which of these answers (literally) feels right?’ is not an effective strategy. Much more sophisticated reactions will need to be stored in order to reconstruct an experience from it affective components. For recall to be successful it is important to establish as broad a frame of reference as possible – i.e. as many affectively significant dimensions as possible, so that the person can react on many dimensions. For example, recalling a Shakespeare play may be aided by relating the characters’ concerns with familiar modern-day dilemmas.

This affective mechanism employed in recall also accounts for the ‘tip of the tongue’ phenomenon where, for example, one can ‘feel’ Paris – the sights, the sounds, etc. – but be unable to reconstruct the specific sounds required to name it.

In every case though, there is a more general conclusion to be drawn – learning would proceed far more effectively if we were able to immerse the person in the experiences themselves, rather than relying heavily on descriptions of experiences. The latter is likely to result in 'token' learning - i.e. factual recall which is not likely to transfer to performance in real contexts. The context-dependant character of memory results from the difficulty in recalling similar affective reactions against a backdrop of quite different emotions – in extreme cases anxiety or other strong emotions may cause our minds to go ‘completely blank’.  In addition, different individuals will have different reactions to the same experience (whether the experience itself or the word describing it) – so whether the approach is recognition or recall, it’s essential to understand the individual and what they care about for encoding to take place.