Flow connectivity in the hillslope–riparian–stream (HRS) system describes the hydrological linkage between upland water and the channel network. However, the time and form of the establishment of HRS connectivity are not adequately understood. Herein, we examined how hillslope structure (topography and soil) and rainfall influence HRS connectivity in a steep, forested, zero-order catchment at the Hemuqiao Hydrological Experimental Station in Southeast China, from July 2016 to November 2017. To this end, surface and subsurface flow, soil moisture, and soil hydraulic conductivity (Ks) were observed, and soil dye staining experiments were conducted. Two patterns of HRS connectivity, namely saturation connectivity that initiates at the soil–bedrock interface (SCSB) and saturation connectivity at different soil horizons (SCSH), were identified. The persistence time of SCSH connectivity was short (<1 h), and the contribution of the perched interflow in different soil horizons to the total runoff was relatively small (0.6–3.0%). Instead, we found that the soil–bedrock interface acted as an important impeding layer that established HRS connectivity. That is, among the rainfall events during which HRS connectivity was established, 90% were established through SCSB connectivity and only 10% were established through SCSH connectivity. We further evaluated the time required to established HRS connectivity and found that SCSB connectivity required more time (4.6–67.4 h) than SCSH connectivity (<1 h). We further found that rainfall intensity determined the initiation of connectivity and that the time required for HRS connectivity decreased exponentially with increasing rainfall intensity (R2 = 0.67). Finally, we found that subsurface saturation excess flow, rather than Hortonian overland flow, was the main contributor to the flood peak during large events. In these events, the total volume of the runoff and flood peak were five and eight times higher than that of subsurface outflow, respectively. These results provide a clearer understanding of runoff generation and can narrow the gap between experiments and models for further development of hydrological theories and methods. © 2020 Elsevier B.V.