“…While there are general agreements about the early (the computation of motion from V1 to MT/V5) and late (the neural representations of 3-D objects at ventral occipitotemporal region) stages of SFM processing – similar to the field of static object recognition – what is largely unknown is the neural representations and computations at the intermediate levels, i.e., the neural mechanisms transferring 3-D motion to 3-D structure. The location of these potential intermediate neurons are also controversial – studies have proposed candidate regions like superior lateral occipital region (sLO) (Kau et al, 2013; Murray, Olshausen, & Woods, 2003; Peuskens et al, 2004, see also data from patients with lesion, Matheson & McMullen, 2010), ventral part of MT/V5 (or human analogues of hMT+) (Kolster, Peeters, & Orban, 2010; Kourtzi, Bülthoff, Erb, & Grodd, 2002; Mysore, Vogels, Raiguel, Todd, & Orban, 2010; Peuskens et al, 2004), anterior superior temporal polysensory area (STPa) (Anderson & Siegel, 2005), and parts of parietal region (Durand et al, 2007; Kau et al, 2013). To reconcile with these different results, Orban and colleagues have proposed a two-stage model of 3-D SFM processing (Mysore et al, 2010; Orban, 2011), the first step involves the extraction of linear gradients at MT/V5, and the second step involves the extraction of second-order gradients at FST, followed by further projections to regions like STPa, IPS, and sLO (Orban, 2011).…”