The mammalian superior colliculus is a midbrain structure importantly involved in eye and head movements that direct gaze toward objects of interest (McPeek and Keller, 2004). This function depends on multiple sources of information. Inputs include direct retinal projections, ascending auditory and somatosensory projections, and contributions from a number of cortical areas (Huerta and Harting, 1984; Kaas and Huerta, 1988; Harting, 2004). As the number and types of subdivisions of cortex vary across mammalian taxa (Merzenich and Kaas, 1980), the sources of cortical projections to the superior colliculus also vary. In addition, as cortical areas have different functional roles, their relevance to superior colliculus functions should vary. Thus, the proportional contributions of different cortical areas must vary within a species.
While there is considerable evidence that a number of cortical areas across a range of studied mammals project to the superior colliculus, there is yet little understanding of the relative strengths of these projections. One reason for this is that most studies of cortical projections to the superior colliculus have been based on placing injections of tracers into subdivisions of cortex. The projections of cortical areas are generally studied one by one. In this way, projections can be demonstrated, but it is very difficult to judge the relative contributions of different cortical areas. While injections of retrograde tracers into the superior colliculus have been used to provide a more global overview of the projection pattern to the superior colliculus, this approach has shortcomings as well. As cortical projections terminate at different depths in the superior colliculus, with caudal visual areas terminating more superficially and rostral visuomotor areas more deeply (Graham et al., 1979; Harting et al., 1992; Lock et al., 2003), injections at different depths in the superior colliculus favor some cortical areas over others in labeling corticotectal neurons. Another complication is that the superior colliculus contains a visuomotor map, and injections of various sizes and locations in the map impact differently in cortical areas of precise or crude visuotopic organization. In both types of studies, cortical areas can be difficult to delimit so that labeled corticotectal cells often can be assigned to specific cortical areas with only limited levels of confidence.
While it is important to recognize these difficulties, quantitative comparisons of projection strengths across cortical areas and across species would seem to depend on studies using retrograde labeling of corticotectal neurons. Using this approach, labeled neurons can be assigned to cortical areas or regions, counted, and the numbers compared across areas to provide a quantitative measure of relative connection strengths. Surprisingly, this has not been done, although there is good evidence that the densities of labeled neurons do vary considerably across the cortical sheet after such injections. Most notably, in an early study in macaque monkeys, Fries (1984) injected horseradish peroxidase (HRP) into the superior colliculus and plotted the locations of labeled neurons throughout the labeled hemisphere. The results across cases clearly indicated that large regions of neocortex contributed little to the superior colliculus, and that much of the input came from caudal visual areas. In addition, deeper injections labeled more cells in posterior parietal cortex and frontal cortex. Relative cell densities were determined for some regions of cortex and found to be similar, but relative projection strengths for cortical areas were not determined. As the study was completed before current understandings of visual cortex organization emerged, visual cortex was only divided into areas 17, 18, and 19. More recently, this type of study was repeated in macaques, and distributions of labeled cells were displayed in unfolded surface views of cerebral cortex, with estimates of the locations and boundaries of currently proposed visual areas (Lock et al., 2003). In this recent study, single injections in each colliculus of three monkeys labeled large numbers of neurons in visual areas V1, V2, V3, and MT. Several other visual areas contained few or no labeled neurons (MST, VIP, 7a, MIP, TE). However, proportions of neurons in various visual areas were not determined. In addition, the uncertainties that exist about the boundaries and even the validity of a number of the proposed extrastriate visual areas (Kaas, 1997) remain a problem.
The present study was directed toward providing a quantitative overview of the areal distribution of corticotectal projections in New World monkeys. While injections in several visual or visuomotor areas have demonstrated cortical projections to the superior colliculus in New World monkeys (Graham et al., 1979; Tigges and Tigges, 1981; Huerta et al., 1986), nothing is known about the total cortical projection pattern. As three species of small New World monkeys have been used in recent studies of visual cortex organization in our laboratory (Lyon and Kaas, 2002), and cortical organization has been studied extensively in two of these species (owl monkeys and marmosets), we included a marmoset, owl monkey, and titi monkey in the present study. Rather than including 2–4 individuals of each species in the study, we injected 2–4 distinguishable tracers into the superior colliculus of each individual. While results varied across injections, total results from 10 injections provided a comprehensive view of the projection pattern. Although this approach has limitations, it addressed our goal of providing an overview for these monkeys.
As an important component of the present study, cortex was separated from the rest of the brain, manually flattened, and cut parallel to the surface. Some of the sections were processed for cytochrome oxidase or myelin, as these markers allowed several areas of cortex to be accurately outlined. The boundaries of other areas could be closely estimated by reference to the histologically defined areas and other landmarks. In this way, we were able to locate labeled neurons accurately in flattened views of cortex, assign many of them to specific areas of cortex with certainty, and attribute others to proposed areas and functional regions with some confidence.