Human learning on the linguistic level is superior to other kinds of learning. Learning at the linguistic level involves the use of language as a medium. In this paper, the human learning is characterized by a process under a natural language environment, and an approach of learning based on indirect linguistic instructions is discussed. An instruction is interpreted through some meaning elements and each trend. Fuzzy evaluation rules are constructed for the searched meaning element of the given instruction, and the performance of a system to be learned is improved by the evaluation rules. In this paper, we discuss a framework of linguistic instruction based learning: FULLINS (Fuzzy-Learning based on Linguistic Instruction). FULLINS is applied to two examples: the truck backer-upper and the helicopter flight control problems. >