The purpose of the study was to experimentally model summation in convergent multiple control over selection-based verbal behavior. Eight undergraduate students participated. In divergent--convergent tact training, a three-choice match-to-sample task was used to establish two selection-based tacts for each of five pictorial stimuli. One textual stimulus served as positive comparison for two pictorial samples, the end result being that each sample controlled selection of two textual stimuli and each textual stimulus selection was controlled independently by two samples. Pre and posttest trials included either a single pictorial stimulus or a two-stimulus compound as a sample, and all five textual stimuli as comparisons. During posttests, each compound predominantly occasioned selections of the textual stimulus related to both of its elements, demonstrating convergent control. Textual stimuli were more likely to be selected in the presence of compounds containing two controlling elements than in the presence of each controlling element in isolation. This finding provides support for a theoretically important aspect of Skinner's (1957) analysis of verbal behavior.