Objectives: To test the hypothesis that intracardiac collateral connections do not result from observable (resolution 40 microns) preexisting vessels and shear stress modulation is not required for collateral formation. Methods: An LCX ameroid constrictor was placed in 6 pigs and the LCX exposed (sham) in 2. After six weeks the hearts were processed and sliced at 40 μ m thickness (3000 slices) in a fluorescence imaging cryomicrotome. High resolution images (2000 * 2000 pixels) were captured, yielding a 3D reconstruction (40 μ m3 voxels) of the vasculature. With custom algorithms, collaterals were detected between (Bridge) and within (Mid) LCX, LAD and RCA. Results: In shams, collaterals were absent. In stimulated hearts 192 collaterals, median ô111 μ m (22–1049), were detected, of which 65% endocardial (p= 0.004). Figure 1 shows a typical case (local opacity altered for visibility). Only 64% of all connections were bridging the LCX to LAD or RCA regions. Mid collaterals (n=24), were found in non-LCX regions with no apparent perturbation of shear stress. Mid collaterals (median 76 μ m, 23–1049) were smaller than bridge (137 μ m, 22–990, Mann-Whitney U p=0.047). Conclusion: Since mid collaterals are parallel to normal vasculature, the novel finding of these vessels contra-indicates a dominant role of shear stress in generating and maintaining collaterals. Higher flow rates are to be expected in Bridge versus Mid connections supporting the hypothesis that shear stress plays a role in outward remodeling. Hence, the detection of substantial numbers of collaterals in regions without flow limitation, suggests additional pathways contributing to collateral growth.