The purpose of this study was to investigate the effects of types of headings on text learning. The main hypotheses to be tested were as follows: (1) There will be a significant difference in text learning between types of headings (embedded, marginal, control). (2) There will be a significant interaction effect in text learning between types of headings and levels of academic achievement (high vs low). (3) There will be a significant interaction effect in text learning between types of heading and levels of test item(high vs low). Subjects of this study were 120 elementary school students (sixth grade) selected randomly on the basis of levels of student's acadmic achievement. And they were randomly assigned to three experimental conditions. The stimulus materials were expository text used by Moon, Seon-Mo and Kim, Young-Seon(1989). It was expository text about "Jupiter is a examiner of Life" and unlearned contents at sixth grade. Each text was consisted of 5 passages and about 2500 words with embedded headings or marginal headings. Testing materials were consisted of 10 items (5 passages x 2 items) of knowledge level and 10 items of conprehension level. To test the hypotheses, 3(heading types: embedded, marginal, control) x 2(academic achievement levels : high vs low) x (2)(test item levels : high vs low) ANONA was applied for text learning. The main findings of the study were as follows : (1) There was a significant main effect of types of headings on text learning. Embedded headings and marginal headings showed higher scores than control condition, and marginal headings showed higher scores than embedded headings. (2) There was no significant interaction effect between types of headings and levels of academic achievement on text learning. (3) There was no significant interaction of effect between types of headings and testing levels of item on text learning.