Efferent axons from area V2 to the middle temporal area (MT) were anterogradely labeled by Phaseolus vulgaris-leucoagglutinin (PHA-L) or biocytin and analyzed in serial reconstructions. Five of seven reconstructed axons had three arbors (each < or = 200 microns in diameter) in layers 3-4, separated by 200-600 microns. Two axons terminated in what was apparently a single focus in layers 3-4. Of 15 additional single arbors analyzed, 12 were concentrated in layers 3-4, and measured 200-250 microns across at their widest point. Three of these arbors were more columnar in shape (about 400 microns in diameter), and extended from layer 4 toward layer 1. This system differs in several features from MT-projecting axons originating from V1. Namely, V2 axons terminating in MT are thinner (approximately 1.0 microns vs. 3.0 microns), their terminal specializations are more delicate, and their arbors are concentrated in layer 4 and overlying layer 3, with no collaterals to layer 6. These differences may reflect the distinctive neuronal populations giving rise to these two connectional systems (different sizes of pyramidal neurons in layer 3 of V2, and a mix of pyramidal and spiny stellate cells in area V1). Differences may have implications for timing factors; that is, impulses from V1, subserved by large-caliber axons, may arrive in MT coincidentally with indirect connections via V2 to MT. Another consideration may be the functional architecture of MT. Regularly spaced clusters of neurons have been described in MT which have similar directionality preferences. The interarbor spacing of cortical efferents is consistent with a columnar organization, but the laminar specificity may indicate recruitment of different combinations of postsynaptic populations by V1 or V2 terminations.