This study compared a grammar program using a linguistic approach with one using a traditional approach to determine how each affects children's ability (1) to construct sentences which have variety in structure, and (2) to comprehend silent reading. Five experimental and five control classes of fourth-grade children were randomly selected from different schools in comparable communities in Florida. The control groups received instruction from materials with an essentially traditional approach, and the experimental group followed materials from Robert L. Allen's "A Linguistic Approach to Writing, Discovery 1" and "Discovery 2." Pre- and post-test scores were obtained from an analysis of writing samples and from the "Gates-MacGinitie Reading Tests, Survey D." All data were analyzed by individual change scores. T-scores were calculated for (1) changes by all classes, before and after, (2) changes between pairs of experimental and control classes, and (3) changes between all experimental and all control children. Results favored significantly at the less than .001 level (1) the experimental group in variety in sentence structure, and (2) the control group in reading comprehension. (Author/JS)