Research

Our research focuses on four broad areas, all involving language, social behavior, and personality or individual differences:

• Coordination in dialog: This line of research is mainly about language style matching, or LSM. We think that when people match each other's language style, or function word use, it indicates that they're socially engaged (i.e., paying attention to each other's words and mental states). In most situations, matching is good—we like people who are similar to us and easy to understand. Style matching predicts positive outcomes in speed-dating, work collaborations, friendly correspondence, published poetry, IM chats, and so on. However, there is also evidence that focusing on your partner rather than the task at hand (e.g., negotiation issues) leads to negative outcomes such as impasse and poor problem solving in competitive conversations.
• Individual language use, health behavior, and personality: Most of this research concerns self-talk (coaching yourself throughout the day or writing about a personal problem). How does language use during self-talk relate to self-regulation and self-improvement? Can we improve people's ability to change their own behavior with subtle writing manipulations? Does self-deception in self-talk ("you will certainly be able to finish 3 papers in one day") behave like other kinds of deception, in terms of linguistic indicators and cognitive-affective consequences? How do traits like neuroticism or impulsivity affect the outcomes of self-talk? We think self-talk is fascinating. It may be the single most ubiquitous and influential source of self-regulation that we have, but it hasn't received much systematic attention in mainstream psychology.
• Community language use and health: The increasing availability of Big Data (loosely defined as datasets that are too big to open in Excel without several minutes or explosions) gleaned from the internet has been a godsend for computerized text analysts in psychology and elsewhere. People increasingly live large chunks of their daily lives online, leaving behavioral trails wherever they go. We can use language use on Twitter to predict disease and understand factors that quicken and slow its spread. Amazon reviews can be used to understand the nature of expertise and decision-making. Facebook statuses can be used to identify factors that constrain or magnify behavioral indicators of personality. The Internet is a goldmine, and we're just starting to be able to analyze its data properly. We think this is an excellent time for psychologists to find a friendly computer scientist and start several collaborations. The work We've done in this area so far, for example, would not have been possible without the help of generous friends from the University of Pennsylvania's Positive Psychology Center, such as Andy Schwartz.
• Literature: Literature is an extraordinarily rich data source. It records how experts at thinking about other people tend to think about other people. In stream-of-consciousness writing, we get a glimpse of how the mind works, or how authors believe it works. Most of the time, we find that art imitates life: Fictional beggars who become kings adopt higher status language, the language of an impending breakup is the same in published poetry as it is in real-life couple's everday conversations, and so on. Currently we're studying how individual differences such as sex and empathy relate to individuals' tendencies to write fictional men and women as similar or distinct (based on function word frequencies). So far we've found that across both expert writers and naive participants, women tend to write characters that are fairly androgynous and men write characters that follow typical gender norms for language (with women being more self-focused and social, and men being more formal and socially distant). This differences is largely explained by sex differences in empathy and systemizing (self-reported interest in people vs. interest in rule-based systems).

Projects

Gendered Language Styles: osf.io/963gp

In this project, we manipulated the language of text prompts to be more feminine or more masculine in style and topic, as well as the labeled sex of the author. Participants read these prompts and rated the author. Across experiments we found that, independent of labeled sex, authors who wrote about relational topics in a masculine style were perceived as lower in relative socioeconomic status than those writing about the same topics in a feminine style. When authors wrote about life course topics, their language style did not affect their perceived status.

Gender Identity Threat: osf.io/k4bjr

This is an ongoing project which is investigating the way people respond to threats to their gender identity, drawing from the concept of precarious manhood. In our first study, we gave participants false feedback about how feminine and masculine they were (which was either threatening or affirming), and asked them to write advice to a friend. We also asked how they would vote on a few current political issues (such as right to refuse service and sex-based restroom restrictions). Initial results indicate that women use more feminine language after receiving gender affirming feedback. Men showed no linguistic response to feedback, but men who were threatened indicated more support for conservative policies than affirmed men, or women in either condition.

Killer's Autobiographies: osf.io/vwq9p

In this project, we compared the autobiographies of Elliot Rodger (a spree killer) and Adolf Hitler with a range of other autobiographies. There were a few intuitive differences—such as killers using more anger and death words—but these seemed more related to the content of their texts than something about their psychology. Along the same lines, the Rodger text seemed to depict relatively typical developmental patterns, which might be unexpected given their highly atypical actions.