By Gordon Rugg
Why should anyone care about schema theory? Well, among other things, it’s at the heart of how society functions, and if you make good use of it, you can become rich, famous and socially successful. That’s a persuasive pair of reasons. This article describes the core concepts in schema theory, discusses some examples of how it gives powerful insights, and relates it to various concepts that complement it.
First, some background. Schema theory was introduced in the 1930s by Sir Fred Bartlett. It’s pronounced like “schemer” which is a frequent cause of confusion if people first encounter the term by hearing it rather than reading it. The core idea is that a schema is a sort of mental template that describes the key features of something. For instance, the schema for a typical car includes having four wheels, a chassis, a body, doors, seats and a steering wheel.
There’s a closely related approach known as script theory. Scripts in this context are a sub-type of schema that describe the key features of an activity – a verb as opposed to a noun. For instance, the script for a pre-arranged dinner at a French-style restaurant includes the actions of booking a table, arriving at the agreed time, being greeted by a member of staff, being shown to your table, etc. We’ll be covering script theory in a later article.
So far, this may sound tidy but not particularly powerful or interesting. When you dig deeper, though, schema theory and script theory turn out to have a lot of uses and implications that aren’t as widely known as they should be. These take us into fields as varied as designing game-changing new products, the law, and measuring novelty in film scripts, as well as the eternal question of why the general public appears collectively unable to have consistent, clear ideas about what it wants. First, we’ll work through the basic concepts.
A simple example: How schema theory makes sense of apparently inconsistent results in market research and surveys
Let’s suppose that you do a survey where you ask a hundred people to describe their dream car, and you find that half of them say it should sit high off the road, while the other half say that it should be as low as possible. That looks like complete disagreement. However, suppose that you instead ask them to describe their dream SUV and then to describe their dream sports car. What sort of responses would you expect to get now?
You’d probably now get a high level of agreement within each description; for instance, the sports car would be consistently expected to be low to the ground, with little passenger space or luggage space.
However, you’d get wide differences between the sports car schema responses and the SUV schema responses. The SUV would be expected to be high off the ground (the opposite of the value for the sports car), with a lot of passenger space (again, opposite to the sports car schema) and luggage space (yet again, the opposite to the sports car).
These two would also be very different from the schema for a small urban car, or the schema for a family saloon.
This is pretty obvious when you’re dealing with vehicles, where the schema for each type is widely and explicitly known. What would happen, however, if you were doing a survey about what people wanted in their ideal house, or their ideal job? Unless you were very careful, there’s a high risk that you’d run into this same problem about having more than one schema involved, and having very different ideals for each schema, which all end up in a confusing tangle if you haven’t kept them separate right from the start.
So, schema theory has very practical implications, especially when you’re trying to gather information from human beings. Before moving on to some applications of schema theory, there are some underpinning concepts that need to be covered.
Concepts and terminology
The plural of schema is either schemata or schemas, depending on personal preference. I usually prefer schemata, mainly because schemas sounds like schemers and therefore risks producing surreal misunderstandings about people scheming and plotting.
Schemata are similar to numerous other concepts, such as categories and genres. However, they’re not the same.
Categories are usually defined by a small set of key features – typically, as small a set as possible, so that the definition is as efficient as possible. Schemata, on the other hand, can vary from very simple to very elaborate. A military designer’s schema for an aircraft carrier, for instance, might be very complex indeed.
Genres are a good example of a concept that works well at an informal level, but which rapidly degenerates into inconsistency and chaos when you try to formalise it, as any music lover will probably tell you. For the purposes of this article, I’ll treat genre as simply another form of category, and not deal with it further here.
The idea of different schemata each corresponding to a different sub-category within a classification has a long history in zoology, where it’s at the heart of formal taxonomy. Each genus, for instance, has its own set of key characteristics, and then each species within that genus will in turn have its own further set of key characteristics that distinguish it from the others in that genus.
There’s a sophisticated and extensive literature on this topic, which overlaps heavily with a similar literature in the field of knowledge representation within Artificial Intelligence (AI). These literatures deal in detail with questions such as whether each species always shares the key characteristics of the higher-level categories to which it belongs (e.g. phylum, class or order). The classic example is whether the higher-level category of bird should have the attribute of being able to fly, since some species of bird can’t fly.
If you’re a new researcher, or an established researcher wanting to bring some new and more powerful tools into your chosen field, then the literature just described may be useful to you – the Artificial Intelligence literature in particular gives you a very rigorous, powerful set of methods and concepts for taxonomy and categorisation. A surprisingly large proportion of researchers in a wide range of fields invent classifications and ontologies without first reading up on the basics of category theory or taxonomic theory. The results are often elaborate and impressive, but also elaborately and impressively wrong, because of not taking into account some key issue that’s well known in the literature on classifications and taxonomies.
Some practical applications
By this stage some readers may be wondering what’s new about schema theory; it looks on the surface like the sort of thing that’s been discussed in philosophy and traditional logic for the last couple of millennia.
The short answer is that it does share a lot of foundations with that work, but schema theory and related concepts such as formal knowledge representation in AI go much further.
For instance, traditional classification had major problems with categories that didn’t have neat, crisp boundaries, such as “old” or “big”. Modern work, in contrast, can bring in approaches such as fuzzy logic and rough set theory, which were developed specifically to deal with such cases, and which enable researchers and developers to handle a wide range of real-world problems in an elegant, powerful way.
Fuzzy logic in particular has become ubiquitous since its invention by Zadeh in the 1960s. In brief, it involves assigning a numeric value to how strongly something belongs in a set – for instance, how strongly an air temperature of 20 degrees Celcius belongs in the set of “hot weather” – on a scale from 0 (not at all) to 1 (completely). It’s simple, but it’s extremely powerful in its ramifications. It’s particularly useful for handling rules such as “if the weather is hot, then change the fuel to air ratio in the engine in this way”. A lot of modern engines have fuzzy logic built into them, to improve their performance, in products as diverse as cars and washing machines.
There’s more about this in our article on categorisation here:
Categorisation and schema theory are an important feature of the law. A lot of legal argument is about the schema to which something belongs, with some schemata being legal, and others being illegal.
A classic example is pictures of naked people. Until recently in the West, pictures of naked people were treated very differently by the law, depending on the schema to which they were assigned. One schema was medical images; another was naturist or nudist images; another was art; another was erotica; another was pornography. Each of these was treated differently by law and by society. Pornography was outlawed until recently in most countries, and even in those countries where it is legal, there are still restrictions on which schemata within pornography are legal. So, a decision about schema membership could make the difference between freedom and gaol.
Fortune and glory
Schemata are also important for product design. An example that’s so familiar we seldom notice it is the desktop metaphor on computer interfaces.
In olden times, people interacted with computers via command lines, with the human and the computer taking it in turns to write text onto the screen. That text wasn’t usually very user-friendly – for instance, it’s not immediately clear to a non-specialist what is meant by the Unix comand rm –r *.* even though its consequences could be very unwelcome (it’s the command to remove all of the files in all of your directories).
Most current software doesn’t work like that. Instead, what you see on the screen is designed to look as much like possible for the schema of an office workplace, with folders containing documents sitting on a desktop, and a waste basket at one side.
The beauty of this approach is twofold. One attraction is that it’s nonverbal, which makes it much easier to use across languages, and easier for people with dyslexia etc to use. Another is that doesn’t require much training, since the user can draw on their existing knowledge of the schema for an office desktop, rather than having to learn a new schema.
That’s a principle that we used for our Search Visualizer software, which draws on the user’s existing knowledge of how documents are structured, and how words are distributed within documents (for instance, key terms are mentioned in the introduction at the start, then often have a section of their own in the main text, where a particular term is used frequently, and then are mentioned again in the conclusion at the end).
The change of schema from the schema of writing text to the schema of the desktop was one that changed the face of computing, and made many people very rich. It has also changed public expectations; people now expect products to be usable without having to read a manual, so having the right schema for the product design can be a major factor in the success or failure of the product.
Drawing on well-established schemata has a lot of advantages. It reduces learning load and cognitive load for the humans involved, and it also makes their knowledge more internally consistent. Social schemata are also useful for smoothing interactions between people, by providing a shared set of norms about behaviour.
However, the fact that a particular schema is well-established doesn’t necessarily mean that the schema is a good thing. A lot of social schemata owe more to history than to logic or evidence or free choice – for instance, the schemata that specify the types of food and clothing and hairstyle that are permissible or forbidden within a society.
The issue of social schemata leads into the concept of the uncanny valley. This concept was first described by Mori, a researcher developing human-like robots. His description of the uncanny valley involves cases where something that looks nearly human, but not quite human, evokes feelings of unease. His description was in the context of robots; other researchers have wondered whether this also explains the unease people feel about other entities in the uncanny valley around humanity – for instance, the legends that many cultures have about half-human, half-animal wild men, and the almost-human monsters in horror movies, such as vampires and zombies and werewolves.
In my book Blind Spot, there’s a discussion of whether this is actually part of a wider phenomenon, in which people feel uneasy about anything that doesn’t fit neatly into one category or another. This unease is at the heart of humour (where we suddenly re-perceive something in a totally different but non-threatening way) and of horror (where some forms of horror involve suddenly re-perceiving something in a totally different and very threatening way). It’s a fascinating topic, to which we’ll return in a later article.
A lot of human activity involves trying to reduce cognitive load by fitting a complicated, untidy world into simple, tidy and familiar schemata. There’s a passage from Buffy the Vampire Slayer in the episode called “Lie to me” which shows the attraction of this approach. The context is that Buffy has just been dealing with a complex, morally difficult challenge that has left her emotionally harrowed and uncertain; she asks her friend and mentor Giles for support. He tells her that life is complex and morally difficult; she tells him that she doesn’t want to hear that. Giles tries a different tack.
Giles: It’s terribly simple. The good guys are stalwart and true. The bad guys are easily distinguished by their pointy horns or black hats and we always defeat them and save the day. Nobody ever dies… and everyone lives happily ever after.”
Buffy: (with weary affection) “Liar.”
Sometimes, when reality is hard going, a simpler version has many attractions. There’s a lot more to be said about schema theory, and about script theory, but that can wait till another day.
Notes and links
The Search Visualizer blog is here:
There’s a free online version of the Search Visualizer software here:
Blind Spot is available on Amazon: