To examine the effect on comprehension of mistakes made while reading, a three-experiment study was conducted. A secondary aim was to determine whether the unskilled reader would make hypotheses about what words in the text should be. In the first test, 80 nine-year-old students with average or above average reading abilities read narrative stories with six types of embedded anomalies, with a view toward both accuracy and the use of context to guess the meaning of the anomaly. In the second test, 10 similar students read stories with fewer anomalies; in the third, 20 unskilled readers, aged 9 to 12, read the same stories without anomalies. Results indicated that the necessity of accurate decoding depended on the desired type of comprehension; accuracy appeared to be necessary for atomistic precision but less so for global interpretation. References, graphs, and a sample instrument are included. (DF)