The present proposal aims to make interspecies comparisons of quantitative models of large-scale cortical systems (macaque, marmoset and mouse). In macaque we have full connectivity data for 27 out of a total 83 cortical areas. Brain-wide high-frequency scanning in retrograde tracer experiments revealed novel findings (Markov et al., 2010; Markov et al., 2011a; Markov et al., 2011b): (i) Cortical pathways are characterized by their weights that span 6 orders of magnitude; (ii) Cortical areas receive 30% more inputs than previously reported. (iii) The cortical graph has a density of 65% (i.e. 65% of the interareal pathways that can exist, do exist). At such high densities the cortical graph cannot qualify as a Small World architecture, contrary to recently proposed models from studies using compiled non-consistent databases. The high density means that a binary description (connected/not connected) of the network fails to capture its specific features. Hence determining the specificity of the cortical graph requires investigation of its physical constrains, given by the weight and distance of its connections. In macaque, this allowed us to show the operation of an exponential distance rule, which we have shown determines the motif distribution as well as the global and local efficiencies in the graph. We propose to: (i) Further develop our data base of the monkey cortex so as to better define its global statistics. We shall make retrograde tracer injections in 23 additional areas and by using improved automated analysis of retrograde tracing we will be able to complete the analysis of the additional 1500 pathways in the course of the project. We shall develop graph-theoretic procedures along the lines developed by Jouve (Jouve et al., 1998) to predict the connectivity of the areas which have not been injected. This information will be used: Firstly, to inform our choice of areas to be injected. Given that physical constraints play a large role in shaping the cortical network. We will therefore play particular attention to the possible existence of specialized components of the network such as hubs and particular motifs. Secondly, to infer the weighted edge-complete cortical graph of all 83 areas, which we shall use to make human macaque comparisons and predict the connectivity of the human brain at a level of detail not possible with existing imaging techniques. (ii) We will examine how brain size influences the statistics of cortical networks. We shall therefore establish a consistent database for the marmoset and mouse cortices analogous to that of the monkey. Given that in macaque cortex weight and distance combined define the specificity of the inter-areal network of the cortex, how do these properties scale with overall brain size, and are the global statistics of the large-scale network primate specific? These questions will be approached by comparing the global statistics of the cortical networks in a large and small brain (macaque vs. marmoset) and comparing primate to non-primate (mouse vs. marmoset). The results across all three species will allow us to define the canonical interareal cortical circuits and to better apprehend important algorithmic features of the network. (iii) We will address whether the statistics of the cortical network change during development. In macaque sensory cortex there is little evidence of the transient developmental connections (i.e. pathways between areas in the young animal that are absent in the adult) that could support alternative physiological properties such as we, and others, have previously shown in kitten. Could it be that in primates transient connectivity is a feature of the prefrontal cortex? Here we will examine the postnatal development of connectivity of a representative prefrontal area, area 10, to see if there are statistical changes of the FLN values of the 55 pathways it forms with areas in the occipital, parietal, temporal, frontal and prefrontal lobes.