The processing of constituent characters in Kanji words in proofreading Japanese sentences

In two proofreading experiments the processing of constituent characters in Japanese words was studied. In Experiment 1, the subjects were asked to detect misspellings on a CRT displaying a character, a word, a phrase, or a full sentence at a time. In Experiment 2, a comprehension task was added to make sure that the subjects read sentences in left-to right fashion with comprehension. Both experiments showed consistent results. The manipulation of visual similarity between misspelled and correct characters revealed that misspellings having supraletter features nearly identical to the correct characters received lower detection rates, and that the subjects utilized word-level information when proofreading. The disruptive effect on character processing in reading a sentence that was expected from the unitization hypothesis (Healy & Drewnowski, 1983) was not found. The constituent-level units in Japanese words are considered to be functionally equivalent to word-level units, instead of constituting a hierarchy.Part of this research was presented at the Second International Conference on Visual Search, Durham, U.K., September 1990