The σ_v–T relation assuming no evolution, i.e. C = 0 in equation (3), for low (left – 0.0 < z < 0.5) and high (right – 0.5 < z < 0.9) redshift samples. The solid blue line shows an orthogonal regression fit to the data with the dashed line representing the 95 per cent confidence interval. The dot–dashed line shows a bisector regression fit to the data (see Section 3.2). A model of the form seen in equation (3) was used in the Metropolis algorithm to determine a line of best fit (see Section 4.1). It is interesting to note that the two best-measured systems (XMMXCS J105659.5−033728.0 and XMMXCS J114023.0+660819.0) in the high-redshift subsample are relatively far off the best-fitting relation, with a higher than predicted temperature. Our current observations do not provide good enough spatial resolution or deep enough multicolour photometry to determine the exact reason for this and require further study and re-observations.