Dreaming improves creative problem solving

In a study just out, researchers at the University of California San Diego tested whether “incubating” a problem allowed a flash of insight, and found it did when people entered a type of sleep known as REM.

When we sleep we pass through different ‘stages’ of sleep that are associated with different types of brain activity. REM sleep is the stage that we know as dreaming. It is detectable by electrical activity in the brain that looks much like the waking state, and by rapid eye movements – hence ‘REM’.

Volunteers who had entered REM or rapid eye movement sleep in this Proceedings of the National Academy of Sciences (PNAS) study were then better able to solve a new problem that required lateral thinking. Lateral thinking involves looking at a problem from unconventional angles in order to solve it, making connections that are not obvious. It involves thinking ‘around’ a problem rather than tackling it head-on.

On the morning of the test day, 77 volunteers were given difficult creative problems to solve. After trying to solve them, they were asked to mull over the problem until the afternoon – either by resting but staying awake, or by taking a nap. The naps were monitored, to check whether or not the participants entered REM sleep.

Compared with quiet rest and non-REM sleep, REM sleep increased the chances of success on the problem-solving task – improving creative problem solving ability by close to 40%.

The results suggest a very special role for dreaming in problem solving. It is not just sleep itself, or the passage of time, that is important for the problem solving, but whether REM sleep has occurred.

The researchers believe that dreaming “creates a richer network of associations for future use”.

Other studies indicate it does more than this: it is actually able to help in unravelling logical connections through unconscious reasoning – an important aspect of fluid intelligence.

Dreaming improves finding logical patterns and seeing ‘the big picture’

A study in 2007 – also published in the PNAS – led by Matthew Walker and colleagues demonstrated that solving fluid intelligence type questions (such as those of the last post), involving abstracting logical relations, and seeing ‘the big picture’ in terms of underlying  patterns, was much improved by sleep.

For instance, if a person learns that A is greater than B and B is greater than C, then she knows those two facts. But embedded within those is a third fact – A is greater than C. This can be deduced by an inference. It is this kind of logical relationship finding that dreaming can also enhance. In other words, dreaming helps with fluid intelligence problem solving.

References

Jeffrey M. Ellenbogen, Peter T. Hu, Jessica D. Payne, Debra Titone, andMatthew P. Walker. (2007). Human relational memory requires time and sleep, PNAS, 104, 7723-7728.

Damian Birney, David Bowman, and Gerry Pallier in their Behavioral and Brain Science commentary on Clancy Blair’s ‘How similar are fluid cognition and general intelligence? A developmental neuroscience perspective on fluid cognition as an aspect of human cognitive ability‘ (2006, BBS, 29, 109-160 – link here) make the following excellent point:

“The gF (fluid intelligence) – working memory – executive function  issue is a further case in point. gF has developed meaning from within the psychometric domain, where it is common to define constructs not only by what they are, but also by what they are not. Hence, using factor-analytic techniques, gF has been empirically defined as the latent trait extracted from a variety of reasoning-dominated tests. This gF trait is related to, but empirically (and therefore theoretically) distinct from, the gC latent trait, which is similarly extracted from various tests of (typically verbal based) acculturated knowledge.

WM (working memory) theory was developed within the cognitive-experimental paradigm, mostly using dual-task methodologies to dissociat various storage and processing systems.

EF (executive function) has a more recent history and has been endorsed most actively by cognitive neuropsychology.

The tasks used in these related, yet distinct, research programs have been developed with different purpose in mind.”

Here is my  model of explicit motivation (as opposed to implicit, introspectively inaccessible motivation), executive processes and fluid and crystallized intelligence. Lots of reviewing of the literature and experiments to do now!

In the New York Times article on this paper (link), the author points out that

Brain-imaging studies of people evaluating anomalies, or working out unsettling dilemmas, show that activity in an area called the anterior cingulate cortex spikes significantly. The more activation is recorded, the greater the motivation or ability to seek and correct errors in the real world, a recent study suggests. “The idea that we may be able to increase that motivation,” said Dr. Inzlicht, a co-author, “is very much worth investigating.”

This is one of those extraordinary ‘connection’ moments for me. This idea relates precisely to the implicit and explicit motivation dual process account I just wrote up in a book chapter. In Dr Inzlicht’s paper, the magnitude of the ‘Error Related Negativity’ ERP component (generated by the anterior cingulate cortex) following errors was significantly correlated with better academic performance as measured by official student transcripts. So here we have a  link between ‘norm violation’ detection and intelligence. Interestingly Dr. Inzlicht seems to understand the effect in terms of motivation: the motivation to correct errors. More on this later.

How might this Anterior Cingulate study relate to Proulx & Heine ’Kafka’ study?  Here is a quote from the paper on meaning compensation:

Several recent studies have suggested that people will also fluidly compensate for meaning threats by affirming unrelated meaning frameworks (e.g., Burris & Rempel, 2004; McGregor et al., 2001; Navarrete, Kurzban, Fessler, & Kirkpatrick, 2004). In response to these studies, we have directly tested and supported the hypothesis that the meaning frameworks people affirm in meaning-maintenance efforts are radically substitutable, such that one meaning framework (e.g., moral beliefs) or another meaning framework (e.g., group affiliation) may be called upon when an unrelated meaning framework (e.g., a perceptual schema) is violated (Proulx & Heine, 2008).

My proposal is that this ‘radical substitutability’ may reflect an underlying normativity learning & self regulating system, that relates to intelligence. When there are threats to meaning (i.e. normativity), this system has a widespread motivational effect on all normatively regulated cognition from perception and grammar all the way up to moral and existential value-systems.