RESEARCH

Solving problems with insight

One area of interest is creative cognition. In particular, one type of creative cognition is exemplified by "insight" solutions to problems. These are the solutions that are accompanied by "Aha" or "Eureka" experiences. It's as if a light turns on and you suddenly see an answer to a problem that had stumped you. Sometimes this happens when you didn't even know you were thinking about the problem.

A series of behavioral and neuroimaging experiments has begun to de-mystify this process, and reveal how the brain produces insight.

PLoS Biology recently published two experiments performed in collaboration with Ed Bowden here at NU, and John Kounios at Drexel University. We gave study participants a series of word problems to solve. Subjects got three words like:

For each problem, the solution is a single word that can form a compound word or phrase for each of the words (e.g., "mind" or "piece" could both work with game, but neither works for all 3; the actual solution is at the bottom of this page). Sometimes people solve these with insight (Aha!) and sometimes through straightforward solving methods. We let subjects (after training) tell us how they solved each problem. While they solved 124 (or more) of these problems, we recorded brain activity with fMRI or EEG, with two separate groups of subjects. Both experiments pointed to important involvement of the same brain area: The anterior Superior Temporal Gyrus of the right hemisphere. No insight effect was observed anywhere within the temporal lobe of the left hemisphere.

We had predicted this area might be involved because it seems also to be important for drawing distantly related information together when comprehending complex language. This is just what is needed to overcome impasse and solve a problem with insight. This fMRI result is also consistent with previous results from our lab demonstrating RH advantages in solution priming (fast responses to solution words) and solution decisions, for words presented to the RH, via the left visual field, compared to words presented directly to the LH, via the right visual field (Beeman & Bowden, 2000; Bowden & Beeman, 1998; Bowden & Jung-Beeman, 2000; Jung-Beeman & Bowden 2003).

The EEG experiment provided two additional pieces of information. First of all, the RH temporal lobe activity appeared as a sudden burst of high-frequency (gamma band) activity, relative to solutions achieved without insight. This neural activity is often associated with complex cognitive processing, in particular binding elements of a percept together as it comes into a consciousness.

A second, unexpected EEG effect also was observed: About 1.5 seconds prior to insight solutions, an increase in lower frequency (alpha band) activity appeared over the right posterior cortex. This effect disappeared precisely when the high-frequency activity began over the right temporal lobe. This may reflect "gating," or attenuation, of visual input, allowing initially weak solution-related activity to gain strength, then burst into consciousness as an insight. This is like closing your eyes so you can concentrate when you are trying to solve a difficult problem but in this case, your brain is blocking out just the visual inputs to your right hemisphere.

We are extending this work with different types of problems, different tasks that allow us to assess component processes, and with assessment and induction of different moods to examine how these moods alter neural systems and cognitive processing.

The answer to the example problem? Card (face card, card table, card game).

Preparing to solve with insight - or with noninsight processing

We have also recently demonstrated, in a paper "in press" at Psychological Science, that a person's brain state prior to each problem influences whether they end up solving that problem with insight or with noninsight mechanisms. In the "preparation period" prior to seeing a problem that they solve with insight, people show increased activity in anterior cingulate cortex (ACC), posterior cingulate cortex (PCC), and left and right posterior temporal lobes (compared to prep periods prior to problems they solve with noninsight processing).   Insight preparation could be characterized as being prepared to strongly activate prepotent candidate solutions while also being prepared to switch attention to nonprepotent candidates, thereby enabling a person to retrieve weakly activated solutions characterized by remote associations among problem elements.  In contrast, noninsight preparation could be characterized as external attentional focus on the source of the imminent problem.  The fact that subjects use both of these forms of preparation suggests either that they spontaneously alternate strategies, or that one form of preparation, presumably insight preparation with its top-down component, is perhaps too cognitively demanding to use for every problem in a series.  

Future studies will further specify these preparation-related brain-states and their determinants.  Ideally, this line of research could lead to the development of techniques for facilitating or interfering with insight in order to optimize performance for different types of problems and contexts. 

Other papers in the works show that such brain states are modulated by mood (positive mood increasing activity in ACC and increasing chance of solving with insight; anxiety doing the opposite); and that individuals' brain states before they even begin the experiment are associated with their tendencies to solve with insight or noninsight processing.

Another line of research investigates how people "fill in the gaps" when understanding stories and complex discourse (see this paper "in press" at Cognitive Brain Research). Consider what happens when you hear:

"Mary bought a big package of white paper, then headed off to her cabin. A month later, she gave her husband a new novel."

You probably inferred that Mary wrote the novel. People seem to make inferences like this effortlessly. But clearly, some cognitive - and neural - effort is required. This is reflected in subtle changes in fMRI signal observed while subjects listen to stories that either imply an event - like the passage above - or explicitly state it (... paper, then wrote a book in her cabin...). Despite the fact that study participants are engaged in very demanding processing throughout the story, we see changes in activity when an event can be, or must be inferred. Thus, we are using fMRI as a "covert measure" of cognitive processing, when subjects have no secondary task to perform. The results suggest that the right Superior Temporal Gyrus of the right hemisphere (possibly the same area observed in the insight experiments) is particularly involved in early stages of "filling the gap," by working on difficult semantic integration. Soon, the emphasis shifts to the parallel region of the left hemisphere. Additional neural activity was observed in the Inferior Frontal Gyrus, which we took as an index of "selecting" (or completing) inferences.

The results of this fMRI study perfectly complement prior research showing that patients with brain damage in temporal/parietal and frontal areas of the right hemisphere have difficulty drawing such inferences (Beeman, 1993; Brownell, 1986). They also parallel results of behavioral studies showing that the inferable concept if first "active" (or accessible) in the right hemisphere, and later in the left (Beeman et al 2000).

In a number of ongoing experiments, we continue to parse the behavior of drawing inferences into component processes, and modulate the processing that we think emphasizes right versus left hemisphere semantic processing.

How do people understand complex language? I recently reviewed my approach (Jung-Beeman, 2005). I am interested in how the brain combines the meanings of words so that we can understand the "big picture" of stories and conversations. Some research uses two or three word inputs, or sentences, or whole stories to try to get the big picture here. Again, multiple methods are employed.

One approach I've taken is to examine differences in the way the right and left hemispheres process information, particular in regards to language and problem solving.

Studying hemispheric differences has tremendous potential to reveal critical components of higher level processing, because the two hemispheres share similar gross structure and input and output pathways, yet differ in cognitive processing. A particularly interesting area to study is language processing, for which the left hemisphere has long been thought to be "dominant." Recent empirical and theoretical work suggest that, although the Left hemisphere is better at many straightforward language tasks, both hemispheres process linguistic information, mostly in parallel, at all levels of processing, each computing the input in a unique way, and each contributing to understanding.

I've proposed that semantic information is represented and processed through the summed distributed activity of many thousands of neurons divided over both hemispheres (see, e.g., Beeman, 1998; Beeman et al. 1994). When processing words, the right hemisphere weakly activates large "semantic fields" of related information, including information only distantly related to the input words. This relatively coarse semantic coding accounts for the right hemisphere's linguistic inferiority, as well as its sensitivity to semantic overlap from distantly related words and its subtle contributions to discourse comprehension. In contrast, the left hemisphere strongly activates a narrow semantic field of closely related information. This relatively fine semantic coding accounts for left hemisphere linguistic superiority, as well as its relative insensitivity to distant semantic relations and vulnerability to misinterpretation in certain discourse and problem solving contexts. Subsumed within this broader model are specific theories of how both hemispheres cooperate when people draw connective inferences from discourse (Beeman et al., 2000) and solve verbal insight problems (Beeman & Bowden, 2000; Bowden & Beeman, 1998).

* How mood affects cognitive flexibility, particularly the neural bases of mood (e.g., positive mood, anxiety) effects on insight problem solving, discourse comprehension, and creative cognition

Positive mood is thought to enhance creative problem solving and performance on insight problems. How does this occur, if it does?

* Right hemisphere involvement in recovery from aphasia

* Discourse and semantic processing in Autism