In the present study, we examined how context of instruction and information in the visual array to be described affect spatial information packaging across a range of levels of spatial description. Participants described complex scenes containing 3-D dollhouse furniture across two different arrays (functional vs. nonfunctional arrangements of objects) and across instructional contexts (living room context, furniture showroom context, no context). Knowledge about the visual scene and instructional context both had an impact on spatial descriptions, but separately, and at different levels of granularity. The influence of visual context was particularly striking, with marked differences across conditions at multiple levels of information packaging descriptive trajectories (the order in which objects in the spatial array were described), amount of detail, and explicit mention of atypical object orientation. The importance of visual context as a means of accessing context frames in common ground is discussed.