Freehand mid-air gestures are a promising input method for interacting with wall displays. However, work on mid-air gestures for wall-display interaction has mainly explored what is technically possible, which might not result in gestures that users would prefer. This paper presents a guessability study where 20 participants performed gestures for 25 actions on a three-meter wide display. Based on the resulting 1124 gestures, we describe user-defined mid-air gestures for wall-display interaction and characterize the types of gesture users prefer for this context. The resulting gestures were largely influenced by surface interaction; they tended to be larger and more physically-based than gestures elicited in previous studies using smaller displays.