Error diagnosis in ICALL typically analyzes learner input in an attempt to abstract and identify indicators of the learner's (mis)conceptions of linguistic properties. For written input, this process usually starts with the identification of tokens that will serve as the atomic building blocks of the analysis. In this paper, we discuss the consequences of mismatches between the learner's perception of a given token and the system's interpretation of its linguistic properties. Based on our analysis of the interaction of beginning learners of Portuguese with the ICALL system TAGARELA, we discuss why tokenization and the interpretation of accented characters deserve particular attention in a system used by language learners. On the computational side, we argue that the mismatches arising in such cases can be addressed in a general way by building ICALL systems on an annotation-based natural language processing architecture which monotonically enriches the representation of learner input. (Contains 1 figure and 2 notes.)