A method that is widely used to analyse stellar populations in galaxies is to apply the theoretically derived responses of stellar spectra and line indices to element abundance variations, which are hereafter referred to as response functions. These are applied in a differential way, to base models, in order to generate spectra or indices with different abundance patterns. In this paper, sets of such response functions for three different stellar evolutionary stages are tested with new empirical [Mg/Fe] abundance data for the medium-resolution Isaac Newton Telescope library of empirical spectra (MILES). Recent theoretical models and observations are used to investigate the effects of [Fe/H], [Mg/H] and overall [Z/H] on spectra, via ratios of spectra for similar stars. The global effects of changes in abundance patterns are investigated empirically through direct comparisons of similar stars from MILES, highlighting the impact of abundance effects in the blue part of the spectrum, particularly for lower temperature stars. It is found that the relative behaviour of iron-sensitive line indices are generally well predicted by response functions, whereas Balmer line indices are not. Other indices tend to show large scatter about the predicted mean relations. Implications for element abundance and age studies in stellar populations are discussed and ways forward are suggested to improve the match with the behaviour of spectra and line-strength indices observed in real stars.