readings> how to create a habit

The experiment was simple enough. The subject slid her head into the maul of the brain scanner in the basement of London's Institute of Neurology and tried to keep very still while also ignoring the ominous clankings and whirrings coming from several tons of surrounding machinery. Then with one hand resting on a key pad, she started using trial and error to work out an unknown sequence of eight finger taps.

A tick or cross would flash up on a screen with each key press to tell her when she had got the right finger and so could move on to the next movement in the series. Finally, once all eight were found, her instruction was just to keep drumming out the pattern until it became an unthinking rhythm. By the end of the hour's session, the subject's fingers were skipping through the complex routine almost of their own accord. Like a car driver weaving through traffic on the familiar route home, she had become barely conscious of the mechanics of what she was doing.

A trivial sounding test, for sure. But one with potentially revolutionary implications for our understanding of what it means for something to be "in consciousness". As Bernard Baars, a psychologist at The Wright Institute in California and editor of the journal, Consciousness and Cognitition, exclaims, it is a beautiful paradigm: "You are comparing two apparently identical outputs. Externally, the actions look much the same. But in one case, the actions come with a feeling of being experienced and in the other, the actions are not experienced. So the level of consciousness becomes your experimental variable and how the brain differs between these two states must tell you a lot about the mind."

Certainly for neuroscientists like Oxford University's Dick Passingham, who last summer reported on the finger-tapping experiment carried out using the Institute of Neurology scanner, the differences are proving startlingly easy to see. When subjects are in the learning phase, having to remember what they have so far discovered about a sequence while groping for the next step, their whole brain is alight with the effort. There are a range of high-level planning and memory areas in the forebrain at work, including the prefrontal cortex, anterior cingulate cortex, and the premotor cortex. But some more lowly brain centres are also working overtime, including parts of the basal ganglia, the thalamus, and even the cerebellum - a bulge off the brainstem once thought to be just a motor reflex centre. Vast swathes of the brain light up as if the brain is throwing every vaguely relevant circuit at the task.

Yet within minutes of "getting" the sequence, this wash of activity begins to drain away. The brain-scanning machine which tracks the surges of oxygenated blood to busy parts of the brain starts to show quite a different picture. The job of moving the fingers becomes confined to just a small set of motor areas. It is as if having used the whole brain consciously to establish the individual finger movements, a routine has now been downloaded to form a fixed and autonomous memory - a skeletal crust of habit. The brain has developed a simple template that can produce the same output "as if" it were still going through all the hoops of being consciously aware.

And while it might not seem such a surprise that the brain should be able to automate a motor task, other imaging studies have shown exactly the same result for more intellectual skills such as thinking up verbs to match a list of nouns, learning a route through a maze, or mastering the computer game, Tetris.

The automation of language has been studied intensively by a team led by Marcus Raichle at Washington University in St Louis - although thinking speech to be a highly conscious activity, the group's experiments did not exactly start out as an investigation of habit learning. The initial intention was merely to show which parts of the human brain "did" language. Volunteers were put under a scanner and then to provoke some level of thoughtful activity, they were asked to find verbs to match a series of nouns. For example, seeing the target word "hammer" should suggest a response like "hit".

Right from the start the results were confounding because the team had been expecting just a few isolated language modules to light up during the performance of this task. Yet as with the finger tapping exercise, much of the brain was afire - both wide areas of the frontal cortex and also low-level centres like the basal ganglia and cerebellum. There seemed too much activity for making what seemed a rather specific mental connection.

Then one day a subject voiced the concern that he might not be able to perform the language task fast enough once under the scanner. This was a reasonable fear as the target list had to be presented at the rate of almost a word a second to keep the minds of the subjects fully occupied over several minutes of recording. So the subject was allowed a few run-throughs with the actual list to be used in the experiment as it was reasoned that exactly the same brain circuits should be used every time a person made the same association between two words. However, when the results came back, this subject's brain showed barely a flicker of activity. From there being seemingly too much activation before, now there was stunningly little.

"When we first tumbled to this, it was just a wow kind of reaction," says Raichle. "We simply didn't expect to see these big sweeping changes with things going up and things going down all over the brain for such a basic task."

What Raichle calls the practice effect has shown up in other scanner studies such as Richard Haier's work with subject's learning to play Tetris at the University of California-Irvine and most recently, with experiments by Steve Petersen at Washington University in which subjects learn the path through a drawn maze - a task deliberately designed to complement the language tests by taxing the brain's visuospatial pathways. Trumpeting these findings in a paper to the Philosophical Transactions of the Royal Society, Raichle says that enough results have now accumulated for their importance to an understanding of consciousness to be obvious. Although, he admits, exactly how to interpret them is another matter.

One point of agreement among researchers involved in the skill learning experiments is that it is clear that paying focal, effortful attention to something calls the whole brain into action. The brain does not behave like a collection of isolated pathways, each doing their own thing, but as coherent system.

As Passingham notes, there are general-purpose planning centres that seem to be called into play whenever the brain is dealing with any kind of novel or difficult mental situation, whether it be making verbal connections or working out a tap sequence. These areas of the prefrontal cortex and the nearby cingulate cortex seem to help "scaffold" the processing of the moment, acting as a temporary scratchpad for a set of goals, memories and possibilities. They hold together a context of ideas for long enough to guide the more specialist language and motor centres to an appropriate state of output. But the surprising level of activation in lower brain areas like the cerebellum and the basal ganglia suggests that these too are put into exploratory mode, watching and learning from what is going on.

However it is what happens next - the rapid downloading of a skill from bright awareness - that seems the real key to the brain's enormous processing abilities. The neuro-imaging experiments suggest that once the brain finds some optimal way to respond to a certain situation, the wider scaffolding rapidly falls away. The brain no longer needs to carry a running memory of its recent performance or actively maintain a representation of the goal state - how doing the skill should feel. Instead, the brain can let the response be reduced to its bare essence, creating a memory trace in the motor or language areas which then lies in wait to intercept a pattern of input, such as the visual stimulus of the word "hammer", should it ever pass by its way again in the future.

In other words, the brain builds its processing pathways by turning a lifetime's succession of small mental experiments into thick strata of habit. Raichle says this much should be obvious from watching a child develop. A new-born baby has to struggle to master the simplest of tasks. It takes excruciating concentration to bring a wavering hand into contact with a toy. But within weeks, the same child will be making the movement with unthinking skill. The challenge then becomes to combine learnt components of action into novel sequences, such as manoeuvring a hand to reach for a toy placed awkwardly round a corner. As a child accumulates these basic motor and perceptual skills, its brain is gradually freed to concern itself with more sophisticated varieties of thought, such as beginning to imagine the consequences of its actions.

This general picture of shrinking patterns of activation as the brain learns from its own fumbling explorations seems well established. But Passingham says it has only been in the past year or so that experiments have begun to offer some sense of the mechanisms behind such shifts.

One key has been to show that prefrontal cortex involvement is essential for a mental event to feel sharply experienced - a fact long suspected by neuroscientists but always difficult to prove. Passingham says that with scanning, demonstrating this could hardly be easier. In another experiment, he simply asked a group of subjects to pay close attention to the feeling of their now established finger-tapping rhythm. Immediately, areas of the prefrontal and cingulate cortex became active again. And the clincher was that the subject's actual performance grew more ragged, as if their brains were being put back into exploratory mode.

Passingham also showed that the need to employ these high-level areas is probably why there seems only room for one thing in the spotlight of attention during any one moment. He scanned subjects trying to learn a finger sequence and also perform Raichle's verb generation task at the same time. The clash of two novel tasks resulted in a stuttering performance - the subjects had to switch between concentrating on one or the other. But if the finger sequence had already been mastered, and so no longer made demands on the prefrontal cortex, doing both became possible.

More controversial is where the activity goes once it falls out of high-level attention. For some, the scanning data suggests that the brain has two distinct pathways - one set of circuits for dealing with novelty and a second route for automatisms. But to others, this is falling back into the trap of thinking of the brain as a modular system. Instead, they say, the distinction should be between the global and the local. Dealing with novelty requires a full brain response. Then once a suitable pattern of response has been established, activity contracts to whichever brain pathway most directly connects a given sensory input to its matching motor or even mental output.

One researcher who has been studying what this means is Avi Karni of the Weizmann Institute in Israel. Again using a finger tapping task, Karni has concentrated on recording the changes that take place in the primary motor cortex, the final staging post for motor commands to the body. He has found, as the template-refining story would suggest, that there is a rapid shrinkage in activation in this area as if it discovering the most efficient neural representation of the movement sequence. But just to confuse matters, if the subjects then practised the same pattern for a few days, the primary motor cortex activity began to expand again. Karni argues however that this apparently contradictory result fits with other evidence that the brain will increase the number of neurons devoted to coding for a particular kind of sensation or motor output if there is a constant enough demand. So really the shrinkage and then expansion are just successive ways of strengthening the representation of a behaviour at a highly local level.

Yet this still does not explain how a localised crust of habit can then step in to intercept the execution of an action. The suggestion is that the brain comes to be constructed of a bedlam of local routines - small distributed habits of perception and reaction that allow the bulk of each passing moment to be processed swiftly and automatically. So when we are driving a car, the changing of the gears and even the switching of lanes would be done at a preconscious level, with familiar or well-anticipated patterns of input immediately triggering a stereotyped pattern of response.

 It would be only when one of these routines struck a snag - say, the gear knob coming off in our hands or a pot-hole suddenly looming in the road ahead - that we would have to switch to a more global and exploratory form of reaction. So what forms our centre of attention effectively becomes self-selecting - the bit of the moment that has turned out to be least routine. Our layers of habit form a mental filter that lets pass only the novel or the difficult. However the problem with this view of the brain is that there would still seem to have to be some central bottleneck so new input would be sure to find its way to the correct habit stored somewhere in the vast maze of brain pathways.

Ann Graybiel, a neuroscientist at the Massachusetts Institute of Technology, believes she just might well have found the answer. Unlike the other skill researchers, Graybiel's approach is the more traditional one of using electrodes inserted into a monkey's brain to record the responses of individual neurons. For some years she has been investigating the workings of the basal ganglia, a chain of lower brain centres that have long been suspected of being a key player in the automation of complex actions and which are connected to the cortex above by a mysterious set of looping paths.

The oddity is that these loops collect news of what is happening across wide areas of the cortex - both sensory and motor - and then funnel it back to some point in the frontal planning region. Recording from a monkey as it learnt an association such as that a clicking noise signalled the availability of a sip of juice, Graybiel found that in the early stages of training, the basal ganglia cells would fire raggedly to all parts of the event - the sound of the click, the turning of monkey's body, its drinking. But quickly there was a dramatic tightening in activity so that the cells fired only to the salient stimulus - the click.

Graybiel says the position of these basal ganglia cells, straddling the cortex loops, means that they are perfectly placed to observe what patterns of sensory cortex activity eventually trigger a particular motor response, then to short-circuit the cortex's long-winded thought processes by firing themselves the instant they saw this same pattern. Because of the way the basal ganglia connect to individual points in the frontal cortex, this would immediately tug on the relevant motor memory. "Instead of having to consider what to do, the basal ganglia could just cause the brain to emit the behaviour as a single chunk," says Graybiel.

Graybiel adds that because all cortex activity feeds through the bottleneck of the basal ganglia, and not just motor activity, they would presumably do the same service for any form of mental action. The basal ganglia would be responsible even for making much of our thinking and speaking rather routine. Graybiel intends now to find ways of making her monkeys pay fresh attention to the performance of a habit to see if, as expected, this throws the basal ganglia back into exploratory mode. "That wouldn't be surprising because we all know what happens when you start thinking about how you are hitting your forehand in tennis, right?" she laughs.

For researchers like Graybiel and Raichle, these findings are exciting because they are pushing neuroscience towards a more systems-level view of consciousness and the brain. Raichle criticises those like Francis Crick who have sought to locate conscious activity in certain kinds of neural firing codes or within isolated brain areas like the visual cortex. "That's like trying to understand how a symphony orchestra can play Beethoven by worrying about the molecular qualities of a violin string. You've got to look at the whole brain in action," says Raichle.

In the meantime Raichle says there is still much work to be done. He personally is intrigued that certain areas of the brain, such as the insula cortex on the lower face of the frontal lobes, seem regularly to be switched off when the brain is dealing with novelty. One controversial guess is that such areas may be critical to the performance of habits and so would have to be actively suppressed to allow the prefrontal lobes to mount their more exploratory response. "We still need to understand the negative side of this better - what areas of the brain do you have to inhibit to be able to think about things in a novel way," says Raichle. It seems that while at last neuroscientists have a paradigm that gets to grips with conscious behaviour in the brain, this particular story can only get more complicated.

Further reading:

"The neural correlates of consciousness: an analysis of cognitive skill learning" by Marcus E Raichle, Philosophical Transactions of the Royal Society London B, vol 353, p 1889 (1998).

"The time course of changes during motor sequence learning: a whole-brain fMRI study" by Ivan Toni and others, NeuroImage, vol 8, p 50 (1998).

home> back to readings