This is one of a series of reports on the National Longitudinal Study of Mathematical Abilities (NLSMA). This report presents a sequence of analyses that explore the relationships between mathematics achievement and three sets of variables: pupil, teacher, and school for students in grades four and six, (NLSMA Reports Nos. 1, 4, 7, and 9 provide related background material.) Achievement criteria used were computation, structure, and a standardized test. While no extensive attempt was made to interpret the data, several results are clear. Under- and over-achievement varies a great deal according to the criterion used. In general, students using modern texts had better achievement on structure while students using conventional texts had better achievement on computation. Scales measuring pupil attitudes toeard mathematics and (to a lesser extent) self-concept and anxiety scales also tended to relate to under- and over-achievement. Teachers classes varied greatly on the mean residual achievement variable. There was a tendency for background variables to affect results before the attitude scales. For related documents, see ED 044 277 - ED 044 283, ED 044 310, ED 045 447, and SE 016 669 - SE 016 691. (LS)