A study examined whether (1) there are untested indices of difficulty that can be used to enhance the accuracy of readability formulas, (2) the same readability formula can be used to assess the readability of text for mature and immature readers and for good and poor readers, and (3) some one model of the reading process or some synthesis of models presents a better account than others of findings on readability and perception. Subjects, 239 secondary school students, completed questionnaires on which they indicated the extent of their knowledge of and interest in each of 10 topics. Test passages on these topics were first scored to determine the total number of cloze errors; then, a sample was scored to devise a system that could be used to describe the effects of incorrect insertions on the structure and meaning of the text sentence. Results showed that sentence unit integrity tended to be maintained or created at all reading and writing levels. When the topic was familiar and the text became increasingly complex, readers who were having difficulty made an increasing percentage of errors which maintained unit structure and created a sensible alternative reading of the otherwise undeleted unit, but when text was less familiar this did not occur. Results also revealed that the processing of units occurring at the beginning of sentences resulted in the perception of a unit more frequently than the processing of units that occurred later in the sentence. (HOD)