Abstract
Although texts recommend the generation of rich data from interviews, no empirical evidence base exists for achieving this. This study aimed to operationalise richness and to assess which components of the interview (for example, topic, interviewee, question) were predictive. A total of 400 interview questions and their corresponding responses were selected from 10 qualitative studies in the area of health identified from university colleagues and the UK Data Archive database. The analysis used the text analysis program, Linguistic Inquiry and Word Count, and additional rating scales. Richness was operationalised along five dimensions. ‘Length of response’ was predicted by a personal, less specific or positive topic, not being a layperson, later questions, open or double questions; ‘personal richness’ was predicted by being a healthy participant and questions about the past and future; ‘analytical responses’ were predicted by a personal or less specific topic, not being a layperson, later questions, questions relating to insight and causation; ‘action responses’ were predicted by a less specific topic, not being a layperson, being healthy, later and open questions. The model for ‘descriptive richness’ was not significant. Overall, open questions, located later on and framed in the present or past tense, tended to be most predictive of richness. This could inform improvements in interview technique.