If only two rods had gone through Gage's head...

The wonderfully diverse repertoire of animal behaviors is, to a great extent, motivated by a universal desire: to acquire or avoid a particular outcome. Such "goal-directed" behavior is contingent upon learning associations between our actions and the predicted consequences, and, as the name implies, using these associations to direct our behavior. For example, many people drink alcohol because they anticipate enjoying the psychological and physiological results, gamblers place bets based on their expected outcome of the game, and dogs learn tricks because they associate the performance with tasty treats.

A key feature of goal-directed behavior is that it can adapt to new (and unexpected) outcomes caused by changes in the environment. For example, say we learn that a particular route is the fastest and most scenic way to get to work, so we tend to take that route. If construction suddenly begins along that route, that same behavior becomes associated with significant traffic delays and unattractive scenery, which is far less rewarding. Luckily, we have the ability to restructure our mental associations and modify our behavior accordingly, seeking a novel route.

This type of learning, in which our brains are able to respond to altered contingencies, is called "cognitive flexibility." People with deficits in cognitive flexibility, such as those with particular types of brain damage, cannot make such adjustments, and thus the consequences of their actions fail to adequately influence their behavior.

The paradigmatic patient exhibiting such deficits is Phineas Gage, a railroad worker from the mid-1800s. There is an excellent write-up of Phineas's story here; I will offer a brief synopsis: while Phineas was working construction, a 3-foot metal rod (1.25 inches in diameter) blasted through the front of his brain and landed 30 yards away. Phineas survived, but had become, according to his physician, "fitful, irreverent, indulging at times in the grossest profanity (which was not previously his custom), impatient of restraint or advice when it conflicts with his desires, at times pertinaciously obstinate, yet capricious and vacillating, devising many plans of future operation which are no sooner arranged than they are abandoned." In short, he was unable to control his behavior in accordance with the perceived consequences of his actions.

It was concluded, much later, that Phineas's behavior resulted from damage to his prefrontal cortex, which is now associated with a range of high-level cognitive processes, including decision-making, executive function, and working memory. A region of the PFC, the orbitofrontal cortex (OFC), has been implicated particularly in cognitive flexibility. Although it is often regarded as "the" region involved in consequence-motivated behavior, recent research has demonstrated that this role is dependent on its connections with a region called the basolateral amygdala (BLA), an area involved in forming associations between behaviors and outcomes. These two regions are thought to form part of a circuit that underlies goal-directed behavior, but the precise contributions of these regions and the nature of their connections is not well understood.

A group from the University of Maryland, led by Thomas Stalnaker, recently published a paper in Neuron that explored this complicated relationship, using a task called "reversal learning." Basically, rats were trained to poke their noses into ports in response to a particular odor (odor A); this behavior was rewarded with sugar water dispensed into the port. Poking in response to another odor (odor B), however, resulted in a bitter solution (quinine) delivered to the port. Once the rats learn to discriminate the odors and their respective consequences, they will habitually poke in response to odor A, and avoid poking in response to odor B.

At this point, the sly experimenters switch the task: when the rat pokes in response to odor A, expecting a delightfully sweet elixir, it is greeted instead with the sharp, biting taste of quinine. After a bit more poking, the rat will figure out that if it pokes in response to odor B, it can receive the sugar water. Thanks to its cognitive flexibility, it will then rapidly adjust to this novel arrangement of outcomes and poke only in response to B.

It is known that animals with damaged OFCs can learn the initial discrimination task, but take much longer than normal animals to relearn the task once the contingencies are reversed. It is also known that neurons in both the OFC and the BLA are known to fire selectively in response to cues associated with particular rewards; in this example, some neurons (group 1) would fire in response to odor A, while others (group 2) would fire in response to odor B. After the rat learns the reversal, this relationship will switch, and group 1 will fire after odor B, and group 2 after odor A.

Until a few years ago, researchers believed that the OFC was responsible for acquiring and signaling these reversed associations to the BLA. The BLA, encoding the behavior/consequence association, would then modify its association to be congruent with the information from the OFC. More recently, however, this group discovered a key difference in firing between these two areas: neurons in the BLA reverse their firing behavior more rapidly than those in the OFC, indicating that the BLA encodes the reversal before the OFC. This pattern is inconsistent with the original hypothesis.

This group developed a new hypothesis: the activity of neurons in the OFC might encode the expected outcome of a behavior, which can then be compared to an actual outcome encoded by neurons in a different area (or areas). If there is a mismatch between the two, as in the reversal learning task, the OFC communicates this conflict with the BLA, which can then modify its association (by reversing the neuronal firing behavior). Once the animal's behavior changes in response to this new association, the expected outcome changes, thus resulting in a change to the firing behavior of the OFC neurons. This communication allows the animal to rapidly reverse previously acquired discriminations.

If the OFC is damaged, the mismatch is undetected. This allows the association in the BLA to persist and continue to guide behavior, even though it is no longer relevant. It's important to keep in mind that animals with damaged OFCs do eventually learn the reversed contingency, it just takes much longer. This group believed that it is this persistence and inflexibility of the BLA associations which impedes learning the novel discrimination, but other brain areas are eventually able to compensate.

This belief prompts a rather counterintuitive prediction: after damaging the OFC, if you remove the persistent association in the BLA, you will remove the impediment to relearning; i.e. cure the problem caused by OFC damage by damaging the BLA too!

So the group tested four groups of animals: those with OFC damage, BLA damage, OFC and BLA damage, and no area damaged (they damaged the areas by injecting an excitatory neurotoxin specifically into the area of interest). All four groups were able to learn the initial discrimination task, but, as had been shown before, animals with damaged OFCs took twice as long to learn the reversed discrimination. Strikingly in accordance with their hypothesis, animals with damage to both the OFC and BLA could learn the reversed association as well as animals which had not been damaged at all (and as well animals with BLA damage, who show no deficits in reversal learning).

They concluded that the OFC is not directly involved in learning the association, but rather generates representations of expected outcomes, which can be communicated with other regions. The OFC/BLA circuit may be the default pathway recruited for directing behavior, but in their absence the brain can call upon other areas that are capable of supporting reversal learning. Thus, damaging the BLA removes the persisting association, allowing mercenary regions to compensate for the loss.

The critical point is that the connections between the OFC and areas like the BLA are important for cognitive flexibilty, not just one brain area or the other. Brain areas do not work in isolation, but in the context of a network (which includes other regions as well). It is this network, and the cooperation between regions, that encodes behavior-outcome associations, and uses them to motivate and guide behavior.