This thesis looks at two recent developments in phonology, feature geometry and underspecification. Three models for feature geometry (Bernhardt, Halle-Sagey and Clements-Hume) and principles of radical underspecification are studied using previously collected data from child substitutions which reflect normal and unusual substitution processes. The goal of the thesis is to determine the ability of the feature geometry models and underspecification to account for the differences between normal and unusual processes in a principled manner. The findings are that neither feature geometry nor underspecification when used alone provides a complete explanation for the difference between normal and unusual substitution processes. However, by incorporating principles of underspecification into the feature geometry models, the normal and unusual substitutions can be distinguished. Modifications to the feature geometries are recommended based on the findings.