Where You Know That You Know


by Jim Schnabel

September 17, 2010

Tires shriek; a horn sounds, metal crunches into metal, and shards of safety glass spatter the pavement. But what really happened?

For such an event, neuroscientists recognize two kinds of perception: The perception of the event itself, and then the perception of how accurately one has perceived the event. The first involves mere cognition; the second is thought to require metacognition—the brain’s perception and evaluation of its own activity.  A paper published in the Sept. 17 issue of Science advances our understanding of metacognition by finding that people with more accurate estimates of their perceptual performance also have more neurons and better connections at the very front of the brain.

The study should help researchers zero in on the precise brain networks that mediate metacognition. That in turn could result in techniques to boost this faculty, and to measure it objectively—so that one could better determine, for example, how reliable a certain witness’s testimony is in a court case. But metacognition is also of great interest for its close links to the still mysterious phenomenon of awareness.

“The brain mechanism whose performance was measured in this study may turn out also to support sensory awareness,” says Hakwan Lau, a researcher at Columbia University whose laboratory is prominent in this field.

The study comes from the laboratories of Ray Dolan and Geraint Rees at University College London, and was led by Ph.D. student Stephen Fleming and postdoctoral fellow Rimona Weil. The team recruited 32 healthy people and asked them to perform a series of simple visual perception tests. In each test, the person viewed two images with vertical black and white lines and had to choose which image had greater contrast. Immediately following each choice, the subject had to estimate his or her confidence in the choice on a scale from 1 to 6. This introspection of perceptual decision-making was, the researchers believed, an essentially metacognitive task.

Measuring true metacognitive accuracy has long been seen as tricky, among other reasons because one’s effectiveness in judging one’s perceptual accuracy will be  affected by that perceptual accuracy. To control for this, the researchers programmed the computer that ran the experiment to make the visual discrimination task harder for those who were perceptually more accurate, and easier for those who were less accurate. In this way, all subjects ended up having approximately the same score in the visual task.

“We then asked whether these people, despite having the same level of basic perceptual decision performance, differed in their ability to monitor that performance,” says Fleming. “And it turned out that they did. Some people were quite good at knowing when they’d done well or badly, while other people didn’t really have that introspective access.”

What in the brain accounted for this variation in accuracy-estimating ability?  Fleming and his colleagues scanned the volunteers using two kinds of magnetic resonance imaging (MRI), first to measure the volumes of “gray matter”—brain cells—in different regions of the brain, and second to map the density of the “white matter” nerve-fiber bundles that connect different brain regions.

Those subjects who were better able to judge their accuracy turned out to have a greater number of brain cells in the anterior prefrontal cortex, or frontal pole. This proved to be true even after the researchers adjusted the results to take into account the different whole-brain sizes for the subjects. A lesser correlation—which might have been statistically significant if the study had included more subjects—was found between metacognitive performance and the volume of the dorsolateral prefrontal cortex, which is immediately above and to either side of the anterior prefrontal cortex.

Measurements of the nerve fibers in the subjects’ brains, using a form of MRI known as diffusion tensor imaging, showed that the thickness of a tract within the corpus callosum, a large bundle of nerves that spans the brain’s two hemispheres, was also correlated with metacognitive performance. This tract is tightly connected to the anterior prefrontal cortex, so the results suggest that metacognition relates not only the volume of this region, but also its connectedness to other regions.

Previous studies have shown that people with brain damage in forward parts of the prefrontal cortex show less ability to introspect (to reflect on their thoughts and feelings). A recent study by Lau and his colleagues also has shown that the disruption of dorsolateral prefrontal cortex activity with magnetic fields impairs metacognitive ability.

“We only stimulated the dorsolateral prefrontal cortex, but if we had stimulated the frontal pole too, maybe we would have found the same effect,” says Lau. “The frontal pole and the dorsolateral prefrontal cortex are anatomically connected, so it’s possible that the two work together for this kind of process.”

Fleming and Lau and others now plan more studies to determine more precisely whether and how these prefrontal regions perform metacognition. But already, says Lau, their anatomical position makes sense, if they are in fact responsible for metacognition. “The frontal pole in particular receives input from higher cognitive areas rather than early sensory areas,” says Lau. “It is at the top of the information processing hierarchy.”

Finding out where and how metacognition occurs could lead to therapies that boost this ability. It also could lead to a better understanding of metacognition’s relationship to other abilities mediated by this same part of the brain. “Working memory, avoiding distraction, and the ability to switch between different tasks, for example, all depend on the same prefrontal structures,” says Lau.

Lau and Fleming hope that the study of metacognition will lead to a better understanding of consciousness, too. In recent years, experiments designed to reveal the “neural correlates of consciousness” have repeatedly implicated the prefrontal cortex [see “Researchers Narrow Search for Source of Consciousness”]. And while metacognition may not always be something of which we are conscious, the two phenomena are often thought to relate to the same evolved function, self-monitoring, which appears to be more highly developed in humans than in other animals.

“If by consciousness we mean something that we communicate to others, something that reflects an ongoing commentary on our perceptions,” says Fleming, “then that seems to me to be very metacognitive—and in that sense, metacognition would reflect consciousness.”