Research on projectably multi-unit tellings (e.g., stories) has largely focused on their contexts of emergence, beginnings, endings, and uptakes (or lack thereof), rather than on their ‘middles.’ The relatively small literature on such ‘middles’ has focused on different types of responsive behaviors when they do occur (e.g., continuers). However, there is virtually no research on relevance rules that might systematically organize these ‘middles,’ including the production of responsive behaviors (or lack thereof) and the management of intersubjectivity. This thesis describes and defends one such relevance rule: Advisors are strongly accountable for responding – either vocally and/or nonvocally – at each and every complex possible-completion place. This relevance rule provides an inferential framework with which to monitor and manage advisors’ understanding of ‘middle’ units. The method used is conversation analysis – including the analysis of deviant cases – complemented by the coding of data and resultant distributional patterns. Data are dual-camera-videotaped, drop-in, advising sessions conducted in English between 20 non-native-English-speaking international students and native-English-speaking advisors working for a university's Office of International Affairs. Specifically, data involve students’ projectably multi-unit problem presentations (e.g., related to Visa status, course scheduling, international travel, housing, etc.).