This dissertation investigates the influence of recombination upon the evolution of mutation rates, and the properties of mutational load in evolving populations, using the tools of experimental evolution. It is shown, in Chapter 1, that conjugation inhibits the spread of mutator alleles in an experimental bacterial population. This result confirms previous experimental and theoretical findings that mutator alleles spread by remaining in linkage with beneficial mutations. The exchange of genetic material across individuals disrupts this process, and thereby makes it more difficult for mutator alleles to rise in frequency. This phenomenon likely plays an important role in limiting mutation rates in natural populations. In Chapter 2, the fitness cost of mismatch repair deficiency is carefully measured in experimental yeast populations, and found to be substantial (>2%). It is shown that this fitness cost is indirect, attributable to a heavy tail of less fit individuals in the distribution of fitness of the mutator population. Separately, the lethal mutation rate of the mismatch-repair-deficient strain is estimated by observing and tracking thousands of budding events of single cells. The reduction in fitness caused by the presence of less-fit individuals in the mutator population and the excess lethal rate of the mutator strain neatly sum to account for the separately measured fitness cost relative to the wild-type strain. The methods of Chapter 2 are extended in Chapter 3 to produce time-series data from an evolving population. A population of mismatch-repair-deficient yeast is founded from a single cell, in order to begin from a state of no mutational load, and the development of load as the population moves towards mutation-selection balance is measured by estimating the distribution of fitness at several time points. Loads are computed at the early time points, and the methods of approximate Bayesian computation are applied to estimate the deleterious mutation rate and distribution of fitness effects. It is found that the deleterious lethal mutation rate is at least 0.03, and perhaps as high as 0.08, in this strain. These results confirm and augment the findings of Chapter 2, and provide the first-ever experimental demonstration of a population approaching mutation-selection balance.