Semiempirical molecular orbital calculations have been shown to predict double minima potential surfaces which are not confirmed by further experimental or theoretical evidence.1These same techniques predict barriers to internal rotation which are not only in large error but may not even follow the experimental trends.2These discrepancies offer a large decrease in self confidence for the application of these techniques to molecular systems where direct experimental evidence is not available, and these are often the circumstances where one wants a theory to be used. Therefore, our research group is in the process of reparameterizing the procedure INDO3to yield barriers to internal rotation, bond angles, and bond lengths within reliable confidence limits. During this process of determining new parameters we have observed that some sets of parameters will also remove the anomalous energy minima predicted by the original INDO parameters. To illustrate this we consider in this paper the molecule CO2.