Pilots’ manual flying skills have notably reduced due to the increase in flight deck automation over the last decades. This has resulted in a growing concern that today pilots lack the skills to safely and successfully prevent or recover from unexpected aircraft upset events or to take over controls after a sudden transition to manual flying. Although the necessity to reverse this decline in manual flying skills, by developing and implementing additional standards and guidelines for (recurrent) training procedures, is a topic of current interest, additional research is required to be able to implement scientifically substantiated standards to ensure pilots receive sufficient training opportunities to develop, maintain and improve manual flying proficiency. A human-in-the-loop-experiment was conducted to objectively and quantitatively evaluate the acquisition, decay and retention of skill-based manual control behavior in a compensatory dual-axis roll and pitch attitude tracking task. In this study, thirty-eight fully task-naive participants were trained in a fixed-base setting in the Human-Machine Interaction Laboratory at Delft University of Technology and subsequently divided into three groups based on their training performance. Performance of the first group was re-evaluated after a period of non-practice of six months, while the second group was retested at both three and six months after training and skill retention of the third group was measured after two, four and six months. The goal of the experiment was to model the decay curve of skill-based manual control behavior and to determine the reacquisition rate of lost skills compared to their initial acquisition rate. To quantify changes in manual control skills, participants’ control behavior was modeled using quasi-linear human operator models. The results suggest that control skills decay following a negatively accelerating decay curve and that lost skills are reacquired at a higher rate than their initial development rate. However, to construct a decay curve which is able to accurately model skill decay over an extended period of time, a larger-scale experiment should be conducted with a larger number of participants and periods of non-practice ranging from a few hours or days up to several years.