Sentiment, salience, and context: Applying text methods to developmental research using puberty as a case example

Text data are increasingly accessible yet analytic approaches often rely on a single method that may obscure information or nuance. We introduce a complementary set of text methods to address this limitation using puberty as an example. Youth (N = 183, M_age=11.73 years, SD_age=.88) wrote open-ended narratives about puberty experiences and we compared these to a matching set of ChatGPT generated responses to demonstrate source comparison. We applied sentiment analysis, frequency analysis, and pointwise mutual information to examine how puberty was represented across sources. Results across methods highlighted that overall sentiment masked negativity tied to socially evaluated or stigmatized changes and that sources differed in emphasizing or omitting changes. ChatGPT systematically underrepresented physical realities of puberty compared to youth, especially regarding menstruation. Findings underscore that no one method provided a complete account of puberty. Integrating complementary text methods can better support complex developmental questions and interpretations of developmental experiences.