Habit, automaticity, and cognitive controlHumans are endowed with a spectacular capacity for conscious thinking and reasoning. However, just as impressive and perhaps even more important for human survival are our abilities to perform complex routine behaviors with relatively little conscious effort. Most activities that are performed repeatedly become "automatized" in the sense that they can be performed with little or no conscious effort, freeing us up to do other things at the same time. Driving is a good example of this kind of automaticity: Most skilled drivers can carry on multiple tasks (such as drinking coffee or holding a conversation with a passenger) while still maintaining control over the car and successfully getting to their destination. Many of the details of driving are executed with great skill yet never seem to enter consciousness.
Research into the acqusition of skills has led to the proposal that behaviors can be characterized as either "controlled" or "automatic," and has suggested that automatized behaviors cannot be controlled. An oft-cited example of this is the Stroop effect: If a person is asked to name the color of the ink that a word is written in, he or she will be slowed down if the printed word does not match the ink color (see Figure).
The Stroop effect is thought to show that reading is such an automatic skill it cannot be suppressed. Some researchers have inferred from the existence of the Stroop effect that highly trained skills are not under cognitive control. However, there is reason to believe that some aspects of skilled behavior are under a great deal of control. The example of driving suggests a number of aspects of behavior that are highly controlled. The driver's habit is to enter an intersection upon seeing a green traffic light; however, if a child runs out into the intersection, the driver can quickly override her habitual action and stop the car. At the same time, automaticity can also lead to human error. Anyone who drives a car has had the experience of setting out for a given location, only to find himself driving a well-worn route instead, due to just the slightest distraction from his goal. This reflects the fact that overriding habitual behaviors requires cognitive effort, which can be disrupted by distraction.
The research proposed here aims to investigate how automatic behaviors are controlled, and how particular forms of training can lead to increased cognitive control abilities. Our research uses functional magnetic resonance imaging (fMRI) to examine how brain systems are engaged during performance. Because the brain is a highly interactive system, we focus on understanding how the interactions between different regions in the brain change as people learn, and how the success or failure of a particular behavior relates to the interactions between these brain regions. We examine these questions using data analysis methods known as connectivity analyses, which allow us to infer from fMRI data how different brain systems affect one another.
The prefrontal cortex is the part of the brain that is most generally associated with the ability to control one's behavior. People with damage to the frontal lobe often become prisoners of habit: For example, they may exhibit "utilization behavior" wherein any object that is placed in front of them will be grasped and used in an appropriate (but unwanted) action, such as writing with a pencil or putting on a pair of glasses. The prefrontal cortex is thought to override habits when they conflict with our goals, by providing top-down control to other parts of the brain that store habits.
Much of our previous research has been focused on understanding the brain systems involved in learning habits, and this research has identified the basal ganglia as a critical structure for automatization. The basal ganglia are a set of deep brain regions that have long been associated with learning. They are heavily interconnected with the prefrontal cortex, and we propose that the control of habits is primarily driven by interactions between the prefrontal cortex and basal ganglia. A major goal of our project is to understand how these regions interact when people exert control over behavior, and how this interaction changes as habits become automatic.
There are many different ways in which people can exert control over their behavior, and one of the most important is the ability to stop a particular behavior before it occurs. Psychologists have developed a technique known as the "stop-signal" task to examine how well people can stop their behavior. In this task, the subject is told to respond to a particular task on all trials unless they hear a tone, in which case they are supposed to withhold their response. By adjusting when the tone occurs following the task stimulus, we can determine the subject's "point of no return," or the point after which they cannot prevent their response from occurring. We will examine whether automaticity changes one's ability to stop their ongoing behavior. We predict that the ability to stop all behavior upon hearing a tone will not change as a task becomes automatic. However, other aspects of control may be affected by automaticity. For example, we can change the stop-signal task so that instead of withholding their response, the subject instead makes an alternate response. Unlike plain stopping, this task cannot be accomplished by simply shutting down the entire motor system. Instead, the specific response has to be replaced with a different response. This kind of control should involve a different set of brain systems than plain stopping, and we predict that this kind of control will be affected by automaticity.
Another type of cognitive control is seen when people are asked to perform two tasks at once. For example, one might be asked to perform mental arithmetic while driving on a winding road. It is almost always the case that performing two tasks at once exacts a cost in performance (known as dual-task interference), but with practice this cost can be eliminated. In fact, the elimination of dual-task interference is often taken as a measure of automaticity, which accords with our intuition that automatic tasks can be performed without interference by other cognitive demands. We have recently found that training under dual-task conditions also changes the way that the task is learned, such that performance is more efficient if the task is later performed on its own. This finding is consistent with the fact that conditions that result in better long-term learning often result is poorer immediate performance, which has led to the concept of "desirable difficulties" in learning. We will examine how the skills that are learned under dual-task conditions differ from those learned under single-task conditions, in order to understand how effort during learning can lead to better long-term learning.
Another question of major interest is whether people can learn strategies that improve their ability to dual-task that do not rely upon extensive training on the particular tasks that are being performed. There is some evidence that this is possible. In particular, training that teaches people to give varying levels of priority to each of the two tasks improves the ability to perform each task, compared to conditions where people are told throughout training to give equal priority to both tasks. This approach, known as variable-priority training, appears to result in generalized improvements in cognitive control; for example, Israeli Air Force cadets who received ten hours of variable-priority training on a video game were more successful in flight school compared to cadets who did not receive such training. We will examine the changes in brain activity that occur following variable-priority training compared to other forms of training. In particular, we predict that the prefrontal cortex will exhibit greater connectivity with other brain regions during dual-task performance following variable-priority training.
In a world where multiple forms of technology beckon for our attention at any moment, an understanding of how people control their behavior and switch between different tasks is more critical than ever. Our research will provide important new insights into how automaticity, training, and cognitive control are related, and will ground these cognitive insights in neurobiology. The results of these studies could inspire new methods of training which allow a greater degree of control of automatized behavior. By characterizing how automaticity relates to interactions between brain systems, our work will provide novel insights into the neural basis of training, which could be used in the future to help further optimize training methods. Our work will also provide insights into the nature of cognitive control that can help to guide development of treatments for neuropsychiatric disorders that involve reduced cognitive control abilities, including attention deficit/hyperactivity disorder, drug abuse, obsessive-compulsive disorder, and frontal lobe lesions following stroke.
