Newman (1984) examined the causal relations between math self-concept and math achievement in an 8-year longitudinal study using Linear Structural Relations (LISREL) analyses. He concluded that math self-concept did not influence subsequent math achievement. However, the study suffered in that math self-concept was inferred from a single-item scale. Newman addressed this problem, proposed a reasonable solution to it, and based his findings on this solution. However, an alternative solution used in the present reanalysis of his data did support the causal influence of self-concept on subsequent achievement. The purpose of this investigation is to demonstrate the problems associated with single-item indicators in LISREL analyses. The study refutes Newman's conclusion that self-concept has no causal influence on subsequent academic achievement, and argues that it is necessary for such studies to have at least two or more indicators for each latent construct. (Author/JAZ)