Want your novel to succeed? Make it more complex

Members of the computer science department at Stony Brook University in New York say their computer model can determine whether a novel will be successful. Some of their findings were surprising.

A bookend holds bestsellers in place on a bookshelf at the Chicago store The Book Cellar.

Ann Hermes

January 10, 2014

Can researchers predict whether a novel will be successful?

Researchers Vikas Ganjigunt, Ashok Song, and Feng Yejin Choi, all members of Stony Brook University’s Department of Computer Science in New York, say they have created a computer model that can determine how well a novel will do. The study focused on both critical and financial success, using examples including the “Harry Potter” series and the Pulitzer Prize-winning novel “Tinkers” as models of successful books.

The researchers used the number of times a novel was downloaded on Project Gutenberg as well as awards data and Amazon sales information to help set parameters for "success."

According to the results of this study, the key to writing a successful novel is found in the book's language.

“Thinking verbs,” such as “consider” or “remember,” make a book more successful than action verbs like “run” and “blink,” say Ganjigunt, Song, and Choi.

“There exists distinct linguistic patterns shared among successful literature, at least within the same genre, making it possible to build a model with surprisingly high accuracy (up to 84%) in predicting the success of a novel,” they write.

Interestingly, one conclusion Ganjigunt, Song, and Choi arrived at is that what the researchers defined as “readability” has an inverse effect on the novel’s success, writing, “Less successful novels have higher readability compared to more successful ones.”

Why would that be so?

Tracing fentanyl’s path into the US starts at this port. It doesn’t end there.

“We conjecture that the conceptual complexity of highly successful literary work might require syntactic complexity that goes against readability,” they write.

It would be interesting to seek out exceptions to the model proposed by Ganjigunt, Song, and Choi. For instance, how many complex and unreadable novels are critical and financial failures? Plenty, would be our guess.