This study draws on multimodal conversation analysis to emically account for moments in second language (L2) English interaction in which speakers appear to be visualizing text as they talk. One way they do this is by slotting out elements of a turn-in-progress in the air, shifting their hand in a slotting gesture from left to right as they say each word to display to their recipient that they are visualizing certain elements of the turn. In other cases, participants use their fingers to 'write' elements of the turn-in-progress on their palms or in the air. The embodied practices of visualizing a turn component by component as it is formulated therefore make public the temporality of its in situ grammatical production. These multimodally accomplished visualizations also provide the speaker with access to a recalled text that helps them produce the spoken equivalent. The study suggests that English-as-a-foreign language (EFL) learners may therefore support their spoken interaction by visualizing written grammar or lexical items, and that multimodal practices such as the precision-timed deployment of gaze and gesture make a seemingly intrapsychological process like visualization a social matter. The data are taken from a corpus of 94 video-recorded paired discussion tests among EFL learners whose first language (L1) was Japanese.