Mortality, driven by both climate and disturbance legacies, is a key process shaping forest dynamics. Understanding the mortality patterns in primary forests in the absence of severe disturbances provides information on background natural dynamics of a given forest type under ongoing climate change. This can then be compared to mortality rates in severely-disturbed stands. Using a large number of sample plots along a gradient from low to high disturbance, we examined the mortality rates and composition of mortality agents in primary mountain Norway spruce (Picea abies (L.) Karst.) forests on different spatial scales. We evaluated the mortality rates and causes of mortality in 28 stands across a large geographical gradient spanning over 1000 km. We resampled (five-year period) 371 plots (16,287 living trees) in primary Norway spruce forests along the Carpathian mountain chain. The estimated overall annual mortality rate was within the previously reported range of background (ambient) mortality, however, stand-level and plot-level mortality rates varied substantially. Over 18% of plots displayed more than 2% annual mortality and 6% of plots even exceeded 10% per year. Stands in the Western Carpathians showed the highest variability in the mortality rate, with 30% of the stands in this region showing annual mortality rates over 5%. At the plot level, mixed-severity disturbances increased variability of mortality rates within most localities. Overall mortality was evenly distributed among size classes up to 50 cm diameter at breast height (DBH). However, the distributions differ for individual mortality agents. Mortality modes were classified into six categories (broken crown, broken stem, uprooted, competition, bark beetle/fungi, climatic extremes). Bark beetle (Ips typographus L.) infestation was the most frequent mortality agent in all stands, whereas the influence of competition as a mortality agent varied substantially. Mortality from abiotically-caused physical damage was similar to that from competition, yet the distribution among modes of physical damage (uprooted, crown, or stem breakage) varied. The lack of clear evidence of mortality agents in some locations implies that many tree deaths are caused by a combination of contributing factors. The results suggest the role of bark beetle as a mortality agent does not equate to severe mortality at large scales. Prevalence of different size classes affected by individual mortality agents underline the high complexity of the mortality process in primary forests.