Tested was the hypothesis that a sequential, linguistically-oriented program of instruction in written sentence structure and in the relationships between written and oral sentences will produce significantly greater achievement in sentence construction than the traditional English program. Nine classrooms of students (grades 3-5) participated in the experimental program, while another group equivalent in sample size, IQ, sex, and socioeconomic background comprised the control group. Control group curriculum contained aspects of traditional grammar and selected usage and writing activities, while the experimental program consisted of the linguistic-oriented "Roberts English Series"; inductive, sentence-building exercises designed by the investigator; and learning strategies based on investigations of English sentences. Evaluated were the student's ability to construct sentences from randomly selected words, to recognize sentence units in nonpunctuated paragraphs, and to determine morphological and punctuation errors in context. Also determined were mean T-Unit lengths and subordination indices. Data results significantly favored the experimental groups at all grade levels. (Author/JMC)