The authors adapted an experimental design to examine effects of instruction prior to entry into a children's museum exhibit on caregiver-child interactions and children's learning. One hundred twenty-one children (mean age = 6.6 years) and their caregivers were randomly assigned to 1 of 5 conditions that varied according to what, if any, preexhibit instruction the dyads received: (a) building and conversation instruction, (b) building instruction only, (c) conversation instruction only, (d) presentation of models of buildings and conversations without instruction, or (e) no instruction or control. Building instruction included information about triangular cross-bracing. Conversation instruction emphasized the use of elaborative wh-questions and associations. When observed in the exhibit, dyads in the groups that received building instruction included more triangles in their structures than those in the other groups. Caregivers provided with conversation instruction asked more wh-questions, made more associations, and engaged in more caregiver-child joint talk compared with those who received building instruction alone. Type of instruction was further linked to differences across conditions in the engineering content of talk, performance during immediate assessments of learning, and children's memory following 1-day and 2-week delays.