A survey of the literature indicates that the Liberal Arts and Education Colleges are the most successful in the use of predictive tests as measures of achievement in college. The magnitude of success is indicated by correlations of about .40 to .70. However, there is a great amount of variation in results from investigation to investigation. According to the writer's findings, there are few studies in prediction of college success in the engineering and science curricula, and the results of these. are inconsistent. The best prediction, though, bi these curricula seemed to come from the Engineering and Physical Science Aptitude Test. The magnitude of success from this may be indicated by correlations which range from .30 to .60. Finally, the miscellaneous items of the literature survey suggested that there may be many non-academic factors influencing academic performance such as health, military, marital, and financial status; in addition several studies indicated that certain personality factors may influence academic performance in college work.

Averages of the predictors, in this study give some initial prediction indication; ranges, however, fluctuate wildly.

The multiple regression of the full set of predictors indicated that somewhat less than one-fourth of those qualities that make successful engineering college work were being measured by the pre-entrance test predictors upon entering college at V.P.I. None of the regression ratios showed significant value. Because of this fact four of the highest values from the first regression were chosen for multiple regression treatment. These data indicate that about one-fifth of the qualities necessary for successful engineering college work were being measured. Furthermore, from these data the V.P.I. mathematics test, high school rank, and I.Q. are significant as shown by the multiple regression ratios for these items.

Estimates of values were made on the regression of selected predictors; the standard error of the estimates confirms the null hypothesis.

Individual correlations for 1958 indicate high school rank, American Council on Education Test, quantitative, and I.Q. have significance as predictors; for 1959 these correlations show predictive significance for V.P.I. mathematics test, high school rank, American Council on Education Test, quantitative, and American Council on Education, total. Low intercorrelations for high school rank and I.Q. seem to indicate predictors that measure different abilities than the other five predictors.

One of the weaknesses of this investigation, one might suspect, as with others of this nature, would be the rather highly subjective nature of the quality credit average, that is, the grading system. Secondly, the sample for this study is entirely too small to permit drawing any but the most tentative conclusions. Furthermore, where there is a slightly significant predictor such as high school rank, from a state school system, the writer is skeptical of correlated values to another state institution. Then one may surmise that present day tests are crude measures of human behavior especially when related to a time period of five years.