In this paper three different residence time distribution modelling methodologies for horizontal subsurface flow constructed wetlands were compared and these were the impulse response, step change derivative and step change integral modelling methodologies. Impulse response and step change tracer studies were conducted on pilot scale horizontal subsurface flow constructed wetlands to generate the concentration-time data to be fed into the impulse response and step change modelling approaches, respectively. For the unplanted reactor, the two step change modelling methodologies suggested the same fluid flow behaviour reflected by almost identical values for the mean residence time (t¯m) and the same value for the number of stirred tanks in series (N) and Peclet number (Pe). The impulse response modelling approach suggested a 7% higher t¯m and a lower degree of dispersion. For the planted system each modelling methodology suggested different fluid flow behaviour. Practical limitations were attempted to be identified for the two types of tracer experiments. The limitations of the experiments could be considered limitations of the modelling methodologies as they depended on the tracer experiments for data generation. We were unable to collect sufficient data on the peak of the impulse response curve for the unplanted reactor and may have affected the impulse response modelling methodology results. Sampling down the length of the reactor revealed that tracer dispersion had an effect of broadening the impulse response curves to the extent that it was almost impossible to identify non-ideal flow behaviour such as short-circuiting. The mathematical techniques employed by each modelling methodology were also critically assessed. We found that varying the size and hence number of subintervals used in Simpson’s 1/3 rule for numerical integration resulted in different values for t¯m/τ for each modelling approach. The lower the sensitivity of the modelling methodology the better as choosing a parameter as arbitrary as subinterval size should not have a noticeable effect on reported hydraulic behaviour. For both reactor systems the step change derivative approach was least sensitive to subinterval selection, reporting a 1% and 4% variation in t¯m/τ for the planted and unplanted system, respectively whereas t¯m/τ determined by the step change integral and impulse response modelling methodologies varied by 10% or more in some cases. We highlighted the differentiation of F(t) to obtain E(t) to be a potential weakness of the step change derivative methodology as it had the capability to amplify background noise which may have affected the calculation of the hydraulic parameters. It was concluded that each modelling methodology had the potential to output a different reactor model for the same reactor and that each approach has its own inherent strengths and weaknesses. The choice of modelling methodology is ultimately dictated by availability of experimental equipment and the designer’s confidence in using each of the respective approaches.