Current conceptual runoff models hypothesize that stormflow generation on the Canadian Shield is a combination of subsurface stormflow and saturation overland flow. This concept was tested during spring runoff in a small (3.3 ha) headwater basin using: (1) isotopic and chemical hydrograph separation and (2) field mapping and direct tracing of saturated areas. Isotopic and chemical hydrograph separation indicated three runoff components: (1) pre-melt subsurface flow; (2) subsurface flow of new (event) water; and (3) direct precipitation on to saturated areas (DPS). During early thaw-freeze cycles, their relative contributions to total flow remained constant (65 per cent, 30 per cent, and 5 per cent respectively). It is hypothesized that lateral flow along the bedrock/mineral soil interface, possibly through macropores, supplied large volumes of subsurface flow (of both old and new water) rapidly to the stream channel. Much higher contributions of DPS were observed during an intensive rain-on-snow event (15 per cent of total flow). Mapping and direct tracing of saturated areas using lithium bromide, suggested that saturated area size was positively correlated to stream discharge but its response lagged behind that of discharge. These observations suggest that the runoff mechanisms, and hence the sources of stream flow, will vary depending on storm characteristics.