Abstract The understanding of flow processes is a key to evaluating treatment efficiency in constructed wetlands. This work focuses on the effects of flow paths on phosphorus (P) and nitrogen (N)retention in four treatment wetlands constructed on pristine peatlands in Finland. Particular attention was paid to water residence time, effective flow area and effective flow depth. Both an artificial tracer test and a new method based on the analysis of stable oxygen and hydrogen isotope distributions were employed. Tracer tests were used to calibrate steady-state flow models created using a groundwater modelling MODFLOW code. Furthermore, concentrations of P, Al and Fe in the peat and concentration of N in the surface water were measured. Surface water tracer distributions showed overland flow to be the dominant flow process and it was divided into a preferential flow area and dead zones. Also, active channel formation was observed during the years of the study (2002–2005). The results indicate that the hydraulic performance might deteriorate drastically within a short period of time. The active flow areas in the peatlands comprised only about 40–48% in summer, meaning that large areas with potential for nutrient removal were left unused. Flow simulations showed that a more optimal length of the distribution ditch will create a larger effective flow area and possibly could prevent channel formation. The peat P concentration was 1.8±3.9 mg g-1, and P was accumulated in the preferential flow area. The peat P concentration correlated positively with Al in the Ruka peatland treating wastewater. The results indicate that precipitation chemicals increase the P retention capacity of peatland substantially and maintain P retention at a stable level despite variable P loads. Furthermore, the results indicate that the accumulation of P to peat via adsorption and chemical precipitation is the major P removal process even after 10 years of loading. In Ruka, calculated N concentrations in surface water obtained with a first-order area model, together with regression analysis of the rate constant, were in good agreement with observed N concentrations. If a removal of 70% is to be achieved, the NH4-N loading to the peatland should be below 0.10 mg m2 d-1.