This study examined why it is difficult to use recursive processing. Kurland & Pea (1989) took up the recursive call of LOGO procedure and suggested that “embedded recursion” (the recursive call was embedded in between) was more difficult than “tail recursion” (the recursive call came last). In Experiment 1, 10 procedures like LOGO were prepared and tested whether participants (N=36) used procedures correctly or not. There were two types of procedures: 5 tail recursion tasks and 5 embedded recursion tasks. Because participants were not familiar with procedures and recursion, they learned procedure of the recursive call in advance. The result showed that the average of embedded recursion tasks was lower than that of tail recursion tasks. It was confirmed that embedded recursion was more difficult than tail recursion. In Experiment 2, we examined whether the difficulty of embedded recursion is due to embeddedness itself or to embedded recursive structure. There were two types of procedures: 5 embedded recursive tasks and 5 only-embedded tasks. The result showed that the average (N=20) of embedded recursive tasks was lower than that of only-embedded tasks. It turned out that the difficulty of embedded recursion is due to embedded recursive structure.