The robustness of LISREL computer program maximum likelihood estimates under specific conditions of model misspecification and sample size was examined. The population model used in this study contains one exogenous variable; three endogenous variables; and eight indicator variables, two for each latent variable. Conditions of model misspecification included errors of omission of structural paths and simultaneous errors of omission and inclusion. All misspecifications were examined with sample sizes of 100 or 200. In general, the effects of omitting a path from one exogenous variable to another were less serious than were those of omitting a path from one endogenous variable to another. Adding a path from one endogenous variable to another endogenous variable affected the estimation of parameters and standard errors less than any other misspecification. Overall, however, few blanket recommendations as to classes of errors most specifically affecting robustness of parameter estimates could be made. This study confirms that LISREL-type modeling should be undertaken only when there is guiding substantive theory. (SLD)