Using informal, objectives-based, or linguistic methods, three elementary school teachers and three experienced item writers developed criterion-referenced pretests-posttests to accompany a prose passage. Item difficulites were tabulated on the responses of 364 elementary students. The informal-subjective method, used by many achievement test developers, allowed maximum freedom of wording and yielded significantly more difficult items. The objectives-based and linguistic methods, in which the item writer chose the foils (distractors), were susceptible to item writer bias. In contrast, the method which provided maximum control, linguistic-based algorithmic foil method, yielded the items which were easy and not subject to bias. It also created items which were insensitive to the pretest-posttest shift in difficulty. Therefore, algorithmic foil methods are promising because they control item writer differences; more research is needed before reasonable item difficulties and instructional sensitivity can be obtained. Teacher-produced items were more sensitive to instruction than those of experienced writers. It is concluded that item-writer differences are real and that field testing is important to identify these differences. (CP)