We analyzed tag-recovery data to estimate instantaneous fishing (F) and natural mortality (M) rates of four lake whitefish stocks in lakes Michigan and Huron during 2004–2007. We tagged and released 22,452 adult lake whitefish of which 8.7% were subsequently recovered. Annual tag-reporting rates ranged from 17.8% to 56.2%. Tag retention was high for the first 5–6 months after tagging, but tag loss increased substantially thereafter. Nine tag-recovery models were evaluated with respect to whether F and/or M varied among stocks, lakes, or years. There was support for three models based on Akaike information criteria. The best model had yearly and stock-specific estimates of F of 0.03 to 0.79 and lake-specific estimates of M of 0.35 for Lake Michigan and 0.60 for Lake Huron. The second best model had yearly and stock-specific estimates of F of 0.04 to 0.71 and a constant estimate for M of 0.52. The third model had yearly and stock-specific estimates of F of 0.04 to 0.85 and stock-specific estimates of M of 0.32 to 0.67. Model-averaged estimates of F ranged from 0.04 to 0.78 and were substantially different than statistical catch-at-age estimates of F. Model-averaged estimates of M ranged from 0.40 to 0.59 and were greater than estimates obtained from prediction equations, possibly due to sea lamprey-induced mortality. We recommend that tag-recovery estimates of F and M be used as Bayesian priors in future lake whitefish stock assessments to help refine mortality estimates for the stocks.