This paper examined the type of cues used by English-speaking intermediate and advanced learners of Japanese when they map novel verbs onto event representations. In particular, we investigated whether they could use structural information (the number of arguments), case particles and thematic roles, in extracting subcomponents of causative events consisting of an Action subcomponent (e.g., a bear pushes a rabbit) and a Result subcomponent (e.g., a rabbit falls down). We conducted two experiments and found the following results. First, to extract subcomponents, high-proficiency learners could use the same cues as Japanese native speakers; however, when they heard a verb in a transitive frame with two arguments or in a sentence with no arguments, their use of those cues is different from Japanese native speakers. Second, low-proficiency learners basically used only thematic roles as a cues to extract subcomponents, although they could extract the same subcomponent as advanced learners when they heard verb with two arguments.