Given the concern for reliability in essay evaluation and the prospect of "error" variance in its absence, methods to promote interrater reliability in the evaluation of written compositions have been developed. These methods reduce variation in the value systems being applied by readers to texts, either by limiting the group of readers to those who are likely to share value systems to begin with or by encouraging readers toward conformity by making them state explicitly the qualities in texts they will reward with high scores or penalize with low ones. Two positions within critical theory that are relevant to the issue of rater reliability in essay scoring are critical monism (asserting that a text can be said to have only one correct meaning) and critical pluralism (asserting that a text can have more than one correct meaning). According to critical monism, if readers could identify an author's intentions reliably and use relative readability to measure how successfully the intentions were realized through the text, these indexes could be used as a means of certifying readers to evaluate student writing. According to the critical pluralism theory, however, one's sense of an author's intention is always traceable to one's interpretive strategy. And since more than one legitimate interpretive strategy is open to readers, so is more than one sense of an author's intention. Thus, texts can be stabilized by stabilizing readers and by teaching them a common interpretive strategy. The important issue is whether the group's judgments can be taken to represent the judgments of the set of competent readers as a whole. Unless these interpretations are identified, the variance arising from differences in readers's interpretive communities cannot be distinguished from random variance. (HOD)