Idioms form a large part of our knowledge of language. If it should ever bepossible for computers to deal with these expressions, it will need a model forthe representation and processing of idioms. In this dissertation a model isdescribed with categorical grammar as the underlying framework. The writeruses the library research and the field research as the method in accomplishingthis scientific research paper. The writer also browses to the internet in gettingtheory related the study. In field research, the writer uses the quantitativemethod in collecting the data by giving a test to the 30 students from the secondyear student Faculty of Letters Gunadarma University, and later made the resultbecome the data of the research. The data were analyzed per number andconcluded the mastery of the second year students based on percentage of thelevels of understanding, average scores, total scores, total correct, and totalerror answer. After data had been analyzed, the writer finds that the levels ofunderstanding of the second year students in forming the understanding idiomsis in the levels below average or poor ( 60 69 ), with the correct percentage63.33 % and the error percentage 36.67 %.