What Words Are Worth: National Science Foundation Grant Abstracts Indicate Award Funding. (June 2019)

• Record Type: Journal Article
• Authors: Markowitz, David M.
• Is Part Of: Journal of language and social psychology. Volume 38:Number 3(2019)
• Page Start: 264
• Page End: 282
• Publication Date: 2019-06
• Subjects: grant funding -- big text data -- National Science Foundation -- open data -- automated text analysis; Sociolinguistics -- Periodicals; Psycholinguistics -- Periodicals; 401.905
• DOI: 10.1177/0261927X18824859
• Languages: English
• ISSNs: 0261-927X

Abstract

Can word patterns from grant abstracts predict National Science Foundation (NSF) funding? In an analysis of over 7.4 million words covering 19, 569 proposals, this article presents evidence that the writing style of NSF grant abstracts corresponds to the amount of money received for the award. The data describe a clear relationship between word patterns and funding magnitude: Grant abstracts that are longer than the average abstract, contain fewer common words, and are written with more verbal certainty receive more money from the NSF (approximately $372 per one-word increase). While such language patterns correspond to award amount, they largely contradict the NSF's call to communicate science in a plain manner, suggesting an inconsistency between the injunctive norms of the NSF and the descriptive norms of science writing. Broadly, the results support a tradition of research that uses big text data to evaluate social and psychological dynamics.