We consider in-situ observations and results from a global circulation model to study the latitude dependence and probability distribution of polar mesospheric turbulence. A comparison of summer observations at 69° N and 79° N shows that mesospheric turbulence weakens towards the summer pole. Furthermore, these data suggest that at both latitudes in about ~70% of all samples there are non-turbulent altitude bins in the considered altitude range between 70 and 95 km. The remaining 30% with detectable turbulence show an approximately log-normal distribution of dissipation rates. A low-resolution model version with a gravity wave (GW) parameterization explains the observed latitude dependence as a consequence of a downshift of the breaking levels towards the summer pole and an accompanying decay of turbulent heating per unit mass. When we do not use a GW parameterization but employ a high spatial resolution instead to simulate GW effects explicitly, the model predicts a similar latitudinal dependence with weakening turbulence towards the summer pole. In addition, the model also produces a log-normal distribution of dissipation rates. The simulated probability distribution is more narrow than in the observations since the model resolves at most mid-frequency GWs, whereas real turbulence is also excited by smaller-scale disturbances. The GW resolving simulation suggests a weaker tropospheric GW source at polar latitudes as the dominating mechanism for the latitudinal dependence.