It is thought that extrastriate cortical mechanisms underlie global motion analysis because of the large areas over which local motion summation takes place (Burr, Morrone, & Vaina,
1998; Downing & Movshon,
1989). Cells in the middle temporal area (MT) are well suited to this task because they are thought to sum multiple, spatially localized, local motion inputs within much larger receptive fields (Burr et al.,
1998; Downing & Movshon,
1989; Movshon, Adelson, Gizzi, & Newsome,
1985). Moreover, lesions to this area in monkey (Newsome & Paré,
1988) and human (Baker, Hess, & Zihl,
1991) disrupt the ability to encode the direction of global motion. There is also a strong correlation between behavioral performance and cellular responses in this area (Britten, Shadlen, Newsome, & Movshon,
1992) because performance can be modified in a predictable manner by microstimulation of these cortical cells (Salzman, Murasugi, Britten, & Newsome,
1992).
The current model of global motion processing posits a two-stage process (Morrone, Burr, & Vaina,
1995). The first stage involves a contrast-dependent, local motion processing in different parts of the field at a range of different spatial scales. The second stage involves the integration of these local motion signals with early contrast saturation (Edwards, Badcock, & Nishida,
1996). The first stage has been identified with area V1 where directionally selective cells have localized receptive fields of different sizes, tuned to different ranges of spatial frequency with a strong dependence on contrast (Movshon & Newsome,
1996). Area MT has been identified with stage 2, at least for translational motion, since its cells have large receptive fields with multiple subunits and exhibit contrast independence (Fine, Anderson, Boynton, & Dobkins,
2004; Movshon et al.,
1985; Rodman & Albright,
1989).
Our interest is whether sensitivity for global motion processing varies across the visual field. If it does, is this due to regional sensitivity differences in local motion processing in stage 1 of the model or regional differences in how these signals are integrated in stage 2 of the model?
This question cannot be answered without considering the spatial heterogeneity of the visual field evident in early cortical areas (Hubel & Wiesel,
1977). The visual field is heterogeneous in terms of spatial scale, with small receptive fields in the center and much larger receptive fields in the periphery (Sasaki et al.,
2001). It is also heterogeneous in terms of its contrast sensitivity, having higher contrast sensitivity at any particular spatial scale in the center of the visual field (Pointer & Hess,
1989; Robson & Graham,
1981). Regardless of whether spatial scale and/or contrast is preserved at the level of global motion processing (Bex & Dakin,
2002), it will have an important influence on the local motion input to global motion detection when the sensitivity of different parts of the visual field are compared. Indeed, without this information it is not possible to obtain valid estimates of either the regional sensitivity or spatial summation properties of the underlying mechanisms.
Spatial scale and contrast are interdependent aspects of local motion processing identified with early levels of visual processing relevant to stage 1 of the global processing model (Adelson & Bergen,
1985; van Santen & Sperling,
1985; Watson & Ahumada,
1985). Is the global motion dependence on eccentricity solely that expected from the spatial scale and contrast dependence of its local motion input (i.e., stage 1) or is there an additional loss due to global processing per se (i.e., stage 2)? When we answer this question, we also explore the related issue of how global motion processing itself depends on spatial scale.
Previously, to derive the relative low-level and high-level contributions to global motion performance, we have measured the relationship between contrast (abscissa) and global motion sensitivity (ordinate) arguing that any purely low-level contribution would shift this relationship horizontally along the contrast axis, whereas any purely global integration contribution would shift this curve vertically along the ordinate (Aaen-Stockdale, Ledgeway, & Hess,
2007a,
2007b; Hess, Hutchinson, Ledgeway, & Mansouri,
2007; Simmers, Ledgeway, Hess, & McGraw,
2003; Simmers, Ledgeway, Mansouri, Hutchinson, & Hess,
2006). Here because of the importance of spatial scale inhomogeneity considerations, we use spatial frequency narrowband elements and measure the contrast thresholds for global motion detection in central and peripheral vision, and then assess whether any central/peripheral global motion sensitivity difference can be accounted for simply by a contrast metric, in this case contrast relative to detection threshold. We do this for stimuli of different spatial frequency, speed and order (i.e., first vs. second order) (Baker & Hess,
1998).