This thesis is focussed on the investigation of matrices for matrix-assisted diffusion-ordered spectroscopy (MAD). Diffusion-ordered spectroscopy (DOSY) is a family of experiments where the resonances in the chemical shift dimension are further dispersed in an extra dimension according to diffusion coefficient. A typical DOSY spectrum shows one single diffusion coefficient for all the resonances coming from one single species. However, If two or more resonances overlap, the diffusion resolution of the DOSY spectrum is compromised and a spurious diffusion coefficient results, intermediate between the species. In case of signal overlap, the use of more advanced processing methods aids to separate two analytes that differ by at least 30% in diffusion coefficient. In practice, many mixtures contain species of similar diffusion coefficients whose resonances overlap in the chemical shift dimension. The addition of co-solutes can modify the chemical environment (matrix), with which different analytes interact to different extents, and enhance the diffusion resolution of DOSY. However, the addition of co-solutes can risk the benefits of DOSY by increasing the probability of signal overlap. Signal overlap in MAD is avoided by using a 1H NMR-invisible surfactant such as sodium perfluorooctanoate (NaPFO), which has replaced each proton by a fluorine atom. PFO micelles are a tunable matrix which allows the separation of analytes via coulombic interactions by adjusting the pH. Differences in diffusion coefficient in NaPFO solution can be analysed using a modified Lindman's law to model the diffusion coefficient as a function of pH. The model rationalises the binding constants of analytes to PFO micelles with good accuracy, subject to the spectral data quality. Another alternative to resolve diffusion coefficients using the invisible MAD approach is by means of a commercially available alkyl surfactant like cetyltrimethylammonium bromide (CTAB). CTAB in high ionic strength solution forms worm-like micelles whose resonances can be filtered out from the final DOSY spectrum. CTAB worm-like micelles have short transverse relaxation times compared to all of the analytes in the mixture. If a transverse relaxation filter is positioned at the beginning of a standard DOSY pulse sequence, as in PROJECT-Oneshot, the strong CTAB signals vanish and leave behind only the analyte resonances and hence avoid signal overlap. Finally, the use of bovine serum albumin (BSA) as a potential invisible matrix, using a similar approach to CTAB worm-like micelles is investigated, using a relaxation-weighted DOSY pulse sequence to suppress most of the BSA background signal (at a cost in analyte signal to noise ratio). An alternative to suppress most of the BSA background and preserve most of the analyte signal is by means of mild transverse relaxation filtration and spectral editing to obtain an edited DOSY spectrum that shows only the analyte signals. Nonetheless, it is a shame that useful MAD results can only be obtained under a narrow set of conditions: i) different mole ratios BSA: analyte to aid diffusion resolution, ii) mild T2 filtration to improve analyte signal to noise ratio and iii) spectral editing to remove residual BSA background.