The written record leads a double life. It is a vast repository of distinct artifacts, narratives, ideas, histories, and codes. As a corpus, it also holds information about the large-scale patterns that drive culture. Recently, a groundbreaking study in the Proceedings of the National Academy of the Sciences, led by SFI External Professor Marten Scheffer, identified a set of striking patterns that suggest that, in the universe of language, an era framed by sentiment and individuality has been on the rise for decades.
Scheffer and his colleagues, (Ingrid van de Leemput, Els Weinans, both of Wageningen University, and Johan Bollen of Indiana University), analyzed words in the repository of Google Books to assess whether there are significant differences over time in the most frequently used words. They found that from 1850 to 1980 words associated with reason rose systematically, while words associated with emotion declined. From 1980 to the present, however, the trend reversed, showing a rise in emotion-oriented words and a decline in reason-oriented words. In the same period (from 1980 to the present) a parallel trend occurred: a decline from collective words to individualistic words, reflected by the ratio of singular and plural pronouns such as I/we and he/they. The group observed these trends across fiction and non-fiction texts alike, and also found the pattern in the archive of the New York Times.
For Scheffer, as Wageningen University reports, interpreting these patterns is difficult, and “necessarily remains speculative,” since it is not easy to grasp what the patterns in the written record suggest about ways the human psyche interacts with texts. Nonetheless, one possible explanation for the trend from 1850 to 1980 is that “the rapid developments in science and technology and their socio-economic benefits drove a rise in status of the scientific approach, which gradually permeated culture.” By contrast, Scheffer argues, from 1980 on, we may have witnessed a process of disenchantment “as the role of spiritualism dwindled in modernized, bureaucratic, and secularized societies.”
The study has received significant attention, in part because it draws on relatively new methods for accessing the cultural patterns in large-scale text data. As a result, PNAS commissioned a response from one of its reviewers, SFI External Professor Simon DeDeo.
DeDeo, a pioneer in the field of large-scale text analysis, replicated the findings of the study and found similar results. He also remarks on both the methodological challenges and some of the interpretive questions the work raises. Researchers using Google Books for large-scale studies must use caution, DeDeo notes. There are “unknown unknowns that accompany a proprietary database whose detailed composition has never been made public.” He adds, “the data in Google Books is weighted by book, not by how frequently individual books are read, which means that some books get far greater attention than others.”
Ultimately, DeDeo is interested in exploring a deep theoretical problem, which he sees as a fundamental question for the future of psychology: “what is the relationship between the written record and what is in people’s minds?” It’s a difficult challenge, but DeDeo suggests that our studies of language might hold a key. “Rationality is a cognitive process, something we can study in the lab,” says DeDeo. “But it is also a social process, a method of coming to agreement—something that happens, often, with the help of language.”
The exchange between Scheffer and DeDeo ultimately helps refine the methods for this fascinating kind of work, and challenges future researchers to craft strong theoretical foundations for inferences about the relationship between text data, states of mind, and cultural patterns.
Read the paper, "The rise and fall of rationality in language," by Scheffer et. al., in PNAS (December 21, 2021) https://doi.org/10.1073/pnas.2107848118
Read the paper, "Using big data to track major shifts in human cognition," by DeDeo in PNAS (January 25, 2022) https://doi.org/10.1073/pnas.2121300119