By modeling soil as a three-dimensional spatially random material, the reliability of shallow rectangular footings against serviceability limit state failure in the form of excessive settlement can be estimated. The methodology to be used is the 3-d random finite element method (RFEM) which combines finite element analysis with random field theory, where the mean, the standard deviation and the spatial correlation length of the modulus of the underlying soil can be controlled through the input data. The study starts by comparing settlements of rectangular footings against classical solutions for homogeneous deposits and then goes on to investigate the influence of footing aspect ratio on probabilistic settlements. Of particular interest is the relationship between the rectangular footing dimensions and the spatial correlation length of the underlying soil. The results of the studies are presented in probabilistic form, where for given statistics of soil input parameters, the probability of footing settlement exceeding a design criterion is assessed. Earlier studies on adjacent square footings indicated that distributions of settlements and differential settlements can be predicted using the geometric average of the underlying elastic soil modulus field. The current work on rectangular footings represents a further step towards developing a general probabilistic design framework for assessing settlements of shallow footings.