This paper briefly discusses and examines some uses of logarithmic transformation in criminological research and describes both the benefits and drawbacks of employing a logged dependent variable in linear least-squares regression. It is argued that the problems frequently corrected by logarithmic transformation of the dependent variable can and should be corrected by other means when a primary concern of the analyses is the theoretically motivated assessment of nonlinear relationships or interaction effects. To illustrate this, this paper uses demographic and violent crime data for urban neighborhoods and proposes an alternative procedure to log transformation involving the use of weighted least-squares regression, heteroscedasticity consistent standard errors, and diagnostics for influential observations. The potential bias of log transformation for research on nonlinear relationship between neighborhood disadvantage and violent crime is illustrated. However, the analysis is applicable to other areas in sociology and criminology. It is suggested that log transformation is a powerful remedy with powerful side effects deserving of a more thorough justification, as opposed to a quick reference. References and appendix