Linguistic Markers of Self-Representation Reflect Daily Emotional Experiences in an Expressive Writing Study

The words people use provide important insights into their emotional experiences. Several measures for tracking emotions through language have been developed, including sentiment (i.e., the positivity and negativity of language) and I-usage (first-person singular pronoun use, thought to reflect negative self-focus). However, emerging research using large language model embedding techniques has quantified not just if people use “I” but how they use it, yielding a measure of self-representation (i.e., how strongly people see themselves as “good” and “able” rather than “bad” and “unable”). This measure tracks mental health outcomes in psychotherapy. Given that emotions and emotion regulation are intimately connected to how we view ourselves, linguistic self-representations may also offer a means for assessing emotional experiences. Here, we examined the associations between participants’ linguistic self-representations and daily emotional experiences in a 14-day expressive writing study (N = 53). Within-person associations showed that on days when participants’ self-representation embedding scores were closer to “good” or “able,” they reported more positive and less negative emotions. At the between-person level, people with self-representations overall closer to “good” experienced lower negative emotional reactivity to expressive writing. These findings suggest that linguistic markers of self-representation hold promise as a tool for assessing emotional experiences, extending prior work linking it to mental health. Translations of these findings could contribute to earlier identification of changes in emotional functioning and provide new ways of assessing metrics of mental health through everyday language.