The problems arising from treating word and sentence complexity as the direct causes of difficulty in comprehension are surveyed in this paper from the perspective of readability formulas. The basic choices and assumptions made in the development and use of readability formulas are discussed in relation to the larger question of text comprehensibility. When considering how close these formulas come to being accurate and informative predictors of comprehension, it is argued that readability formulas are not the most appropriate measure and that they cannot reliably predict how well individual readers will comprehend particular texts. The paper shows that the high correlations between formula predictions based on text features reported in most readability research are the by-product of using an inappropriate statistical model that aggregates texts and readers, giving an exaggerated impression of the contribution of linguistic factors in the text concerning comprehension. Text and reader properties that cannot be measured by formulas are emphasized as having a far greater influence on comprehension. Further, it is argued that no readability formula can be a reliable guide for editing a text to reduce its difficulty. (Ten pages of references are included.) (JD)