Horizontal instruction techniques used to improve reading skills were tested with 63 limited-English-speaking, Spanish-dominant high school students. The students were divided into four groups. Two experimental groups received instruction in how to infer the main idea of a paragraph through sequential lessons in classification skills. Two control groups were taught how to infer the main idea, but without the use of classification techniques. One experimental and one control group were taught in English and the others were taught bilingually. Contrary to expectation and previous research, results on word classification and paragraph analysis tasks showed no statistically significant evidence to support use of the classification techniques. However, some evidence was found for a treatment effect for abstract paragraph tasks, although it was not statistically significant. The two bilingual groups showed significant pre- to post-test gains, unlike the monolingually taught groups. A variety of methodological limitations could have produced these results. Further research is recommended. (MSE)