The aim of this thesis is to make a contribution toward an eventual characterisation of causal factors in human cognitive development. This is accomplished by providing a rich, microanalytic description of change in a particular skill: seriation. The problem was tackled in two experimental phases, an extended clinical assay and a touch screen based assay. In the extended clinical assay, children aged 3 to 6 years were given the opportunity to review the task, locate errors, and repair them. The main finding was that intervention, for the most part, did not change performance --- subjects under age 4 years were entirely resistant to success, although some older subjects (aged 5 to 6 years) were able to succeed with assistance. In the touch screen based assay, subjects were given intensive training on variations of the seriation task suggested by McGonigle and Chalmers' decomposition of the seriation task. The scope of the thesis was limited to the serial control component of seriation (using arbitrary and non arbitrary strings). Subjects under age 4 years were focused on for intensive training. Despite intensive training under immediate feedback conditions, it was found that subjects continued to be resistant to success. The source of subjects' limitations in reporting items, whether ordered arbitrarily or otherwise, was a limitation of serial control. Without the relevant underlying competence in place, training therefore appears to have only limited effect. That some older subjects can improve their performance indicates that only when the relevant competence is already in place can instruction be effective.