Dual counterstream features	Functional and structural correlates
(1) Source populations of FF and FB pathways are completely separate, which has been recognized as a key requirement enabling distinct functionalities of generative top-down networks (Friston, 2018). A core feature of the dual counterstream architecture is that despite the ubiquity of bifurcation (Kennedy and Bullier, 1985), in no layer do individual neurons in FF and FB pathways possess both up and downstream collaterals (Markov et al., 2014b). Similar findings hold for mouse (Berezovskii et al., 2011). This indicates that FF and FB cells indeed constitute distinct populations.	Molecular characterization of FF and FB neurons is very much on the agenda. This would lead to the development of markers of these two projection types and hold the promise of the development of genetic tools for independent manipulation of different FB pathways.
(2) Large variability in the laminar configuration of FB pathways. FF connections are highly stereotypical, while FB projections show a large variability in their laminar origins and targets. Retrograde tracer in a midlevel target area labels up- and downstream areas with different proportions of supragranular neurons according to their hierarchical distance (Figure 10C). These laminar distributions of FF and FB neurons constitute their signatures that allow areas to be ranked in the Felleman and Van Essen hierarchy. Importantly, the different signatures of hierarchical distance reflect the changing laminar composition of the long-distance inputs to an area, meaning that “hierarchy” refers to the distance-dependent laminar-constrained integration of FF and FB inputs (Barone et al., 2000; Markov et al., 2014b) and signals (Bastos et al., 2015b; Michalareas et al., 2016) into the canonical microcircuit.	The multiplicity of the FB pathways in terms of laminar origins and targets over different distances suggest that compared to the singularity of FF pathways, individual FB pathways come in different flavors and fulfill different functions. These different FB pathways need to be investigated in human and non-human primates (NHP) in high-field, laminar resolution fMRI and in parallel using invasive electrophysiology in NHP. These functional investigations are of particular relevance to the supragranular counterstream given the unique features of the Primate supragranular layers which are generated by a primate-specific germinal zone (OSVZ) (Smart et al., 2002), that exhibits complex cell lineages (Lukaszewicz et al., 2005) that we have hypothesized generate unique cell features in primates (Dehay et al., 2015; Harris and Shepherd, 2015). These considerations support the notion that the two FB pathways (one in the supra- and one in infragranular layers) will have distinct functional roles in generative networks.
(3) FB and FF are not serially organized. Computational modeling studies e.g. (Bastos et al., 2012) invariably assume a serial hierarchical organization (Figure 10C) whereas in fact most areas receive and project to most other areas (Markov et al., 2014b).	The non-serial nature of inter-areal connectivity raises difficulty for ongoing attempts at large-scale computational modeling that needs to be given further attention.
(5) Lamination The dual counterstream architecture explicitly links pathways to layers, which is in-line with the increasing importance attached to cortical lamination and connectivity (Senzai et al., 2019). (l): effects on topography of FB. In addition to the low divergence-convergence values of supragranular layers (L2 FB and L3FF), and the high divergence-convergence values of infragranular layers (L5 FF and L6FB), high-rates of bifurcation by L6/L5 ensures innervation of multiple areas contrasting with and low levels of bifurcation by L2/L3 neurons (Kennedy and Bullier, 1985).	There are two interrelated issues here. Present understating of inter-areal connectivity in NHP, crucially lacks insight into the laminar restricted connectivity of these pathways. This requires using viral tracers in order to obtain precise information on the connectivity of individual pathways over different distances.
(6) Differences of FB and FF targets. The L3 FF projections primarily target upstream interneurons in L4. Presently we do not know the cellular targets of L2 FB projections although there is some suggestion that they could principally target L2 FB neurons in downstream areas, leading to long inter-areal FB chains. In parallel with differences in FF and FB cellular targets, do the cell targets change (and how) as a function of connection distance, possibly bringing a solution to addressing the problem raised in point (3)?	Laminar restricted connectivity to different cell types needs to be investigated in rodents and primates using next generation of viral tracers.
(7) FB show distinct development strategies. Differences in functionality are expected to require different developmental programs, which is indeed the case for FF and FB pathways. FF pathways develop precociously and exhibit directed-growth; FB pathway formation is delayed well into the postnatal period and exhibits diffuse growth followed by pruning (Barone et al., 1995; Barone et al., 1996; Kennedy et al., 1989).	These observations invite the investigation of inter-areal processing in NHP at different developmental periods and suggest that high-field, laminar resolution fMRI in infants and adults could reveal important developmental processes.