There was a recent NIH videocast lecture discussing the ventral v dorsal model of the visual system. My impression is the model was debunked in line with more massively parrale and network models of brain functions.
The ventral vs dorsal dichotomy originates from some differences in object processing correlating with the lesion and functional activation studies, e.g. prosopagnosia with right temporobasal lesion / activation focus and pure alexia in the homologue regions in the left. Positional tasks and spatial tasks comprising tasks with respect to the own body are correlated to dorsal lesions / activations. From a clinical / lesional pov, already the cerebral akinetopsia (lesion at the lateral temporooccipital junction) raise some concern. In my opinion, connections from primary brain areas in a modality to other modalities should be more accounted for, and of course there must be a connection between dorsal and ventral pathways, which makes this dichotomy a good similar kick-off to test hypothesis' against as it was the case in language localisation.