Quantitative MRI (qMRI) is of increasing interest for many neuroscience and neurology applications, motivated by a desire to replace qualitative MR imaging with reproducible and broadly comparable measurements of tissue properties. These measurements are typically made by fitting observations to a model of how the tissue is expected to respond, and hence they depend on the complexity of the model. The MR community as a whole has found it difficult to standardise qMRI and to produce results that are comparable between centres. Our own experience with this began with an attempt to perform multi-component relaxometry on neonates; a failure to produce reliable results led to a broader investigation of rapid gradient echo based relaxometry sequence in general. This work indicated that a significant cause of variability comes ‘magnetisation transfer’ (MT) between the water that is imaged by MRI and protons in semi-solids (i.e. macromolecules) that are abundant in human tissue. I will discuss this issue and present a potential solution that could lead to more reproducible results. I will also discuss methods that we have developed at KCL to model MT effects in MR sequences, and also to measure them. This includes recent work on so-called inhomogeneous MT effects that look to be quite specific to myelinated tissues. Finally, since there is now a large (and increasing) number of different quantitative MRI methods, I will discuss how we could try to evaluate them in order to decide which represents the best use of limited time.