The educational potential of robotics kits as a form of control technology will remain undervalued until meaningful observation parameters are identified to enable a better understanding of children?s control strategies. For this reason, this paper aims primarily to identify and classify the heuristics spontaneously applied by 6-10 year old children interacting with robotic devices containing specific transparency features (i.e. programmability) and interactivity features (i.e. immediacy of feedback). Two studies are described: an exploratory investigation into the control of a Lynx AL5A arm and a pilot study about the control of a Lego Mindstorms NXT®. Two issues relating to control heuristics are addressed: the heuristic shift and the perceived and objective level of task difficulty. The results demonstrate that three main types of heuristic emerge: (i) procedural-oriented, (ii) declarative-oriented, and (iii) metacognitive-oriented. Limitations of the difficulty indicators used and shift patterns proposed are discussed in relation to future research.