Modulation of the flow of Rutford Ice Stream, Antarctica, has been reported previously at semi-diurnal, diurnal, fortnightly and semi-annual periods. A model that includes non-linear response to tidal forcing has been shown to fit closely observations at fortnightly frequencies. Here we examine that model further and test its fit at a larger set of observed frequencies, including the large semi-annual displacement. We show analytically that, when forced by major tidal terms, the model (using a basal shear stress exponent m = 3) predicts several discrete response periods from 4 hours to 0.5 years. We examine a 1.5 year GPS record from Rutford Ice Stream and find that the model, when forced by a numerical tide model, is able to reproduce the reported semi-annual signal. We confirm that about 5% of the mean flow is due solely to the (m = 3) non-linear response to tidally varying basal shear stress. Our best-fitting set of model parameters is similar to those originally reported using a much shorter data record, although with noticeably improved fit, suggesting these parameters are robust. We find that m ≈ 3 fits the data well, but that m ≈ 2 does not. Furthermore, we find that a small variation in flow over the 18.6 year lunar node tide cycle is expected. Fits to semi-diurnal and diurnal terms remain relatively poor, possibly due to viscoelastic effects that are not included in the model and reduced GPS data quality at some discrete periods. For comparison, we predict the response of Bindschadler Ice Stream and Lambert Glacier and show, given identical model parameters, a similar response pattern but with ∼1–2 orders of magnitude smaller variability; these may still be measurable and hence useful in testing the applicability of this model to other locations.