Fifty years ago, Sarnoff Mednick defined the process of creative thinking as
“the forming of associative elements into new combinations which either meet specific requirements or are in some way useful. The more mutually remote the elements of the new combination, the more creative the process or solution.”
Mednick argued that creative people have flat associative hierarchies: they are better able to access distant, remote associations. For example, when given the concept “table”, Mednick predicted that creative people would be more likely to retrieve more remote associations such as “leg” or “food”. In contrast, Mednick argued that less creative thinkers have a steep associative hierarchy, in which words with a higher associative strength (e.g., chair) would be more likely to enter their minds:
Based on his theory, Mednick developed the Remote Associates Test (RAT). The RAT test presents you with three seemingly unrelated words (e.g., “fish-mine-rush”) and you have to find a fourth word (e.g., “gold”) that ties the other three words together.
While the RAT has been criticized as being more related to IQ and working memory than creative cognition, recent research on a number of fronts suggests that Mednick’s theory is sound.*
One exciting area of research is the application of computational network tools to examine the semantic memory network of creative people. Based on mathematical graph theory, a semantic network is comprised of “nodes” (concepts or words) and “links” that indicate the distance between them. Researchers are starting to use the tools of network science to elucidate various aspects of the creative process.
In Melissa Schilling’s network model of cognitive insight, insight can be viewed as the emergence of clarity among a tangled web of thoughts and ideas. According to Schilling, cognitive insight occurs when an atypical association is made, resulting in a shortcut in a person’s network of semantic representations. Insight affects the organization of the entire network, causing a decrease in path length, a new perspective on the entire network, and a cascade of other connections to come online.
Network science methodologies have also been used to compare the writings of prominent poets (e.g., Dylan Thomas) with prominent writers (e.g., F. Scott Fitzergerald). Sarjoun Doumit and colleagues found that poems show a “flatter” associative hierarchy than prose. According to Mednick’s theory, this means that poems are more creative than prose, and involve the combination of more distant associations.
In case you’re wondering what a Fitzgerald network looks like, here’s the network of the 100 most frequent words used in The Great Gatsby (the word size reflects the frequency of use):
In contrast, here’s the much more densely connected network of the 100 most frequent words for the poet Dylan Thomas:
What about everyday creative people? In a recent study, Yoed Kenett and colleagues applied network science tools to directly examine Mednick’s theory among the general population. They administered several creativity measures to a large sample of participants, and based on that information classified participants as either high in creativity or low in creativity. Then they had participants generate free associations to 96 words. They calculated the semantic networks of each group by assessing the overlap of association responses (“associative clouds”) between the words.
If Mednick is right, then creative people should have a richer, better connected, and more flexible associative network than less creative people. And that’s exactly what they found. The semantic memory network of people with low creative ability was much more rigid, spreading out and breaking apart into more sub-parts. In contrast, the semantic memory network of people with high creative ability was much more densely connected (flatter) and thus less rigid, making them more likely to connect associations that are distantly related to each other.
Latent Semantic Analysis
Another tool that has been used recently to look at creative cognition is Latent Semantic Analysis (LSA). LSA an objective scoring method where you can derive a measure of semantic distance. With LSA, you can quantify the similarity between words or texts based on statistical analyses of the responses a large population (e.g., general reading up to a first-year college level).
Using this technique, Ranjani Prabhakaran, Adam Green, and Jeremy Gray found that “thin slices”of verbal behavior (single-word utterances) predicted creative cognition. In their “verb generation” task, a noun was presented every few seconds, and the participant was instructed to say the first verb that came to mind in response to the noun. The researchers analyzed the semantic distance of the responses using LSA. Note that their measure of semantic distance was not a measure of the unusualness of the verb itself, but the unusualness of the verb in the context of the noun (the same nouns were presented to all participants).
They found that the greater the semantic distance of the noun-verb pairs, the higher the levels of creative cognition, story writing ability, openness to experience, and creative achievement. Interestingly, they also found a relationship between semantic distance and traditional measures of “convergent thinking”, such as IQ and working memory. This suggests that creative thinking doesn’t just involve combining distant associations, but also requires consciously accessing the associations. This is where cognitive control–the ability to control thought and action–comes into play.
Along similar lines, Roger Beaty and colleagues recently used LSA to explore individual differences in associative abilities. Participants completed two verbal fluency tasks (e.g., listing synonyms for the word “hot”), and their responses were compared for semantic similarity to the target word (“hot”). They found that large semantic distance values were related to the quality of the responses on a creative cognition task (coming up with alternate uses for a box and a rope). But cognitive control also contributed to creative cognition: Individual differences in IQ and the ability to strategically search memory also played a key role in creative idea production. These results held even after taking into account associative ability.
Why would IQ be related to creative cognition? Paul Silvia and Emily Nusbaum found that the link between IQ and the creativity of responses was related to “executive switching”– the ability to flexibly shift between different mental categories. Additionally, Benedek and colleagues found that IQ was also related to updating: the ability to monitor incoming information and quickly revise the contents of working memory based on the goals of the task.
Interestingly, IQ was not related to inhibition (the suppression of dominant but irrelevant response tendencies), even though both updating and inhibition were related to the creativity of responses on a creative cognition task (e.g., “What can be used for speedy travel?“). Also, note that in the Beaty and colleagues study, associative ability was associated with creative idea production even after taking into account IQ and memory retrieval ability.
Therefore, while controlled cognition is important, creative cognition is more than merely cognitive control. This has important implications for assessing and developing creative ability in students. It also has implications for the creative process.
The Controlled Chaos of Creativity
A clear theme emerges from a look at all of these recent studies. Creative cognition relies on both cognitive control (updating, flexibility, and inhibition) and associative chaos (a loosely structured knowledge base). This is in line with the recent neuroscience of creativity.
Two crucial brain networks for creative thought are the Executive Attention and Default Mode brain networks. The Executive Attention Network is recruited whenever cognitive control is necessary, and involves efficient and reliable communication between the lateral (outer) regions of the prefrontal cortex and areas toward the back (posterior) of the parietal lobe.
In contrast, the Default Mode Network (which I often refer to as the “Imagination Network”) is involved in “constructing dynamic mental simulations based on personal past experiences such as used during remembering, thinking about the future, and generally when imagining alternative perspectives and scenarios to the present.” The Imagination Network involves areas deep inside the prefrontal cortex and temporal lobe (medial regions), along with communication with various outer and inner regions of the parietal cortex. This network draws on regions associated with the retrieval of deeply personal memories as well as semantic information stored in long-term memory. A recent review found that The Imagination Network was highly active during creative thinking.
These two networks can work together to perform certain mental functions. In particular, whenever a task depends on the maintenance or evaluation of internal information, these two brain networks work in concert with each other. Interactions between the Executive Attention and Imagination Networks are especially important for generating new ideas that are both novel and useful.
One of my favorite models of the creative process is the Geneplore model, grounded in the creative cognition approach. According to this model, creativity involves two main processes: generative processes and exploratory processes. Generative processes are important for generating a variety of potentially creative ideas, and includes memory retrieval, accessing distant associations, and analogical transfer. In contrast, exploratory processes are important for evaluating and implementing the most promising ideas. Controlled attention is very important for the exploration of ideas. Over large time scales, these processes will operate in a cyclical fashion, with a constant back and forth between generation and exploration.
I like thinking of creative cognition as controlled chaos because it incorporates both elements of the creative process. Cognition that is merely controlled is rigid and colorless. Cognition that is pure chaos is bizarre. But we owe some of the greatest inventions of all time to cognition that is both controlled and chaotic.
© 2014 Scott Barry Kaufman, All Rights Reserved.
* Although see here for a critique of some aspects of Mednick’s theory.
Acknowledgments: Thanks to Paul Silvia and Yoed Kenett for their assistance with various aspects of this article.
image credit #1: istockphoto; image credit #2: benedek and neubauer; image credit #3 & 4: doumit; image credit #5: pnas; image credit #6: mixingmemory.blogspot.com; image credit #7: flickr user camilo rueda lópez