Discussion of jsiegle's comment: The authors imaged GCaMP3.3 fluorescence in axonal boutons to disambiguate two hypotheses about cortical tuning properties: (1) specific response patterns of higher-order visual regions arise from local processing of untuned inputs or (2) higher-level areas adopt the selectivity patterns of their tuned inputs. Previous work from this and other labs has shown that cells preferring high and low spatial frequencies are intermingled in V1. Cells in the 9 other visual areas, however, can be tuned more specifically. In area PM, the majority of cells respond best to high spatial frequencies (that is: detailed, slow-moving stimuli), while cells in area AL tend to prefer low spatial frequencies (coarse, fast-moving stimuli). These experiments demonstrate that this selectivity is primarily due to the way axons project out of V1, rather than processing within either area. By injecting a viral vector, the authors labeled cell bodies in V1 only, then subsequently imaged their synapses in higher-order areas. The tuning of the synapses matched the average tuning of the cell bodies in each area. In addition, they injected GCaMP into a fourth visual area, LM, and found that its projections to AL and PM were also functionally specific.