We reanalyse observations of Type Ia supernovae (SNe) and Cepheids used in the local determination of the Hubble constant and find strong evidence that SN standardization in the calibration sample (galaxies with observed Cepheids) requires a steeper slope of the colour correction than in the cosmological sample (galaxies in the Hubble flow). The colour correction in the calibration sample is consistent with being entirely due to an extinction correction due to dust with properties similar to those of the Milky Way (R-B approximate to 4.6 +/- 0.4) and there is no evidence for intrinsic scatter in the SN peak magnitudes. An immediate consequence of this finding is that the local measurement of the Hubble constant becomes dependent on the choice of SN reference colour, i.e. the colour of an unreddened SN. Specifically, the Hubble constant inferred from the same observations decreases gradually with the reference colour assumed in the SN standardization. We recover the Hubble constant measured by SH0ES for the standard choice of reference colour (SALT2 colour parameter c = 0), while for a reference colour that coincides with the blue end of the observed SN colour distribution (c approximate to -0.13), the Hubble constant from Planck observations of the cosmic microwave background (CMB) [assuming a flat Lambda cold dark matter (?CDM) cosmological model] is recovered. These results are intriguing in that they may provide an avenue for resolving the Hubble tension. However, since there is no obvious physical basis for the differences in colour corrections in the two SN samples, the origin of these requires further investigation.