Reber Grammar and Implicit Learning

Reber grammar was invented by psychologist Arthur S. Reber in the mid-1960s to serve as a tool in his studies of implicit learning. Implicit learning refers to the acquisition of knowledge about the environment that takes place without the learner's awareness of either the process or the result. We can think of that as unconscious or subliminal learning. Reber grammars are used extensively for research in the psychology of learning in healthy and disordered minds, studies of the effects of brain trauma, functional neurology, and the application of artificial neural networks to the learning of complex sequential tasks, like procedures and natural languages. More recently, the general term 'artificial grammar' has come into common use.

Weird words

A Reber grammar is a network of rules that restrict sequential choices for the letters used to form strings, like words,  There can be many specific grammars, each defined by a different set of rules. Each specific grammar can generate a possibly huge number of words. This kind of thing is also called by other names, such as finite-state grammar, stochastic string generator, finite-state automaton, directed graph, Markov process. I think 'string generator' is the most straightforward way to understand how a Reber grammar works, but 'string' here means a series of symbols, like letters or characters in some language. You start with a set of symbols and then you make a set of rules that are applied sequentially. Each rule restricts what symbols can follow any other particular symbol. The result is something like:

BTSSXXTVVE BPVVE BTXSE BPTTTTVPSE BPTVPSE BTXXTTTVPXVVE

What a Reber grammar does for people who study learning is to define a kind of non-obvious structure that we know is there to be discovered by a learner, one that we can use to test for a learning process. Consider the example result above. You can probably begin to figure out what some of the rules are. Can you identify the symbol set and specify the entire set of rules from the sample data? That would be explicit (conscious) learning. What Reber did, however, was to present his test subjects with a list of these artificial words to study first, without telling him that any rules even applied to them. After their studying time was up, he told them that those words obeyed certain rules (but still didn't tell them what the rules were). He then asked the subjects to look at more words and decide whether each might fit the same rules as the ones they had studied.

"I don't know. It just seems right."

The results showed that the subjects were able to identify conforming words at a level significantly above chance, even though they could not explain just how they decided. They were not consciously aware of the rules. Reber concluded that learning was taking place, but that the learning occurred outside of consciousness and that the rules that were learned were not readily accessible to the learners. Further research using functional MRI has indicated that implicit learning involves a different group of brain areas than explicit learning and memory activities.

So what?

The results seem easy enough to accept. Some might even call it common sense, and I don't think that is far from being right. I would be surprised if you couldn't think of examples of this kind of learning in your own experience. From my experience in learning Chinese characters, I remember being surprised at being able to 'guess' the approximate pronounciation of characters I'd never seen before after having learned a few thousand of them.

Still, Reber's conclusion has met significant resistance in the academic fields of psychology and education. There are important questions concerning just what is being learned and why the success rates are so low. I think these questions suggest a broader range of questions about what the objects of learning processes (the things learned) really are and what it means to learn something. When we think of learning, we think of rules, but maybe much of what we include in the concept of learning is not a process of rule acquisition at all, but a much foggier process that ends with improved ability to guess, to form hunches and gut feelings, or to intuit rather than the ability to know and follow rules. That avenue of thought is in synch with stochastic learning processes such as machine learning techniques and Bayesian belief networks.

The implications of implicit learning range far and deep, so the phenomenon is worthy of rigorous study and analysis. Education is the most obvious area that should be concerned with implicit learning, as traditional approaches strongly favor explicit learning techniques and conscious cognitive involvement. The idea of unconscious learning also touches the roots of individual responsibility and the liability of purveyors of 'hidden persuaders'. There is even at least one attempt to rethink aesthetics on the basis of implicit learning.

 

 


 

  • A.S. Reber. "Implicit learning of artificial grammars", Journal of Verbal Learning and Verbal Behavior, 6, 1967.
  • O'Reilly and Munakata, Computational Explorations in Cognitive Neuroscience, MIT Press, 2000.
  • Skosnik and Reber, "The Neural Substrate of Artificial Grammar Learning"
  • Journal report on fMRI study on implicit learning
  • Reber string generator
  • R. Reber, Schwartz and Winkelman, "Processing Fluency and Aesthetic Pleasure:
  • Is Beauty in the Perceiver's Processing Experience?" Personality and Social Psychology Review, Vol. 8, No. 4, 2004.

Log in or register to write something here or to contact authors.