The association between nurse staf?ng and patient violence on psychiatric units is unclear, and the association between nursing skill mix and violence has not been studied. Monthly data on physical/sexual assaults by patients against others were obtained for 351 adult psychiatric units. Total and injury assault rates were modeled as dependent variables using hierarchical Poisson regression, with total staf?ng and registered nurse (RN) mix as predictors. Assault rates were 12% higher per additional nursing hour per patient day but 6% lower per 5-point increase in percent of hours provided by RNs. In cubic spline models ?t to explore nonlinear staffing�violence associations, assault rates increased with staffing through most of its range but began to decline at very high levels.