[en] One aim of this thesis is to point out the advantages of quantifying (estimating) NMR model parameters directly in the measurement (time) domain, rather than to first transform the data to the frequency domain and carry out the quantification there. Advantages mentioned are: 1. The distribution of sampling times may be arbitrary. It suffices to provide a list of those times at which data points are available. In principle one can calculate the distribution that yields the best overall result. 2. Point 1. implies that truncation of data points does not entail corrections. This in turn implies that the disturbing influence of a broad, non-descript background line can be reduced simply by discarding initial data points. In addition, a measurement may be ended before the signal has decayed into the noise. The latter is important for reducing the time of multidimensional measurements. 3. If the available data points have been sampled at uniformly spaced times, then black-box quantification methods that save human involvement can be used. If, in addition, the exponentially damped sinusoid model function can be entertained, then the computational cost of interactive time domain quantification can be reduced by a factor of several tens. A number of applications of both black-box and interactive quantification methods has been given. The results supported the above points 2 and 3. Subsequently, there are two aspects common to both the time and the frequency domain. First, significant improvement can be achieved by imposing prior knowledge. In this thesis prior knowledge of the phases of the sinusoids was imposed. Second, the quantification may be hampered by lack of knowledge about the model function. In the time domain this problem can be alleviated by expanding an unknown form into a sum of exponentially damped sinusoids with the aid of HSVD. In the frequency domain a partial solution may be achieved by integration of an unknown shape. Note however that use of so-called 'natural lineshapes', derived from high-SNR reference data, may be a viable alternative in some circumstances. (author). 137 refs.; 19 figs.; 7 tabs