This paper presents a method for estimating the primary current distribution in the brain from measured neuromagnetic fields, called minimum-norm estimates (MNEs). The method is based on estimation theory and does not assume any specific form for the current distribution, except that it is spatially restricted. Simulation experiments show that the method can accurately describe the structure of current flow. By increasing the number of measurements, the estimate can be made more localized. The current distributions can also be used to interpolate and extrapolate measured field patterns. The method is applied to the inverse problem of magnetoencephalography (MEG), where the primary-current distribution is estimated using a linear combination of magnetometer lead fields. The lead field describes the sensitivity of a magnetometer to the primary current. The method is based on the idea that the primary current distribution can be represented as a vector in a function space, and the minimum-norm estimate is the current distribution with the smallest norm that reproduces the measured signals. The method is also applicable to electroencephalography (EEG) and has been applied to the inverse problem in electro- and magneto-cardiography (ECG and MCG). The method is robust to noise and can be regularized to avoid numerical instability. Regularization involves suppressing directions in the current space that are poorly coupled to the sensors. The regularized minimum-norm estimate is used to estimate the primary current distribution. The method is also used to create isocontour maps of MEG data, which show the spatial pattern of the data at selected time instants. These maps are useful for visualizing the data and comparing results between different laboratories and sessions. The method is convenient and reliable for neuromagnetic data interpolation and extrapolation.This paper presents a method for estimating the primary current distribution in the brain from measured neuromagnetic fields, called minimum-norm estimates (MNEs). The method is based on estimation theory and does not assume any specific form for the current distribution, except that it is spatially restricted. Simulation experiments show that the method can accurately describe the structure of current flow. By increasing the number of measurements, the estimate can be made more localized. The current distributions can also be used to interpolate and extrapolate measured field patterns. The method is applied to the inverse problem of magnetoencephalography (MEG), where the primary-current distribution is estimated using a linear combination of magnetometer lead fields. The lead field describes the sensitivity of a magnetometer to the primary current. The method is based on the idea that the primary current distribution can be represented as a vector in a function space, and the minimum-norm estimate is the current distribution with the smallest norm that reproduces the measured signals. The method is also applicable to electroencephalography (EEG) and has been applied to the inverse problem in electro- and magneto-cardiography (ECG and MCG). The method is robust to noise and can be regularized to avoid numerical instability. Regularization involves suppressing directions in the current space that are poorly coupled to the sensors. The regularized minimum-norm estimate is used to estimate the primary current distribution. The method is also used to create isocontour maps of MEG data, which show the spatial pattern of the data at selected time instants. These maps are useful for visualizing the data and comparing results between different laboratories and sessions. The method is convenient and reliable for neuromagnetic data interpolation and extrapolation.