Gradient of the linear fit between TMP and X_c/X_i vs the centre of each slice in log(P). Negative gradients imply that more evolved stars (lower values of X_c/X_i) have higher values of TMP in that log(P) slice, as we expect from tidal theory. The gradient is normalised by the mean value of the TMP for each slice, preventing gradient bias due to the larger range of TMP values present at lower orbital periods. The equivalent figure without this orbital-period dependent TMP normalisation is available in the online supplementary material online, along with an animation illustrating the fit for each log(P) slice.