In many species, certain life-history stages are difficult or impossible to observe directly, hampering management. Often more easily observed stages are monitored instead, but the extent to which various forms of uncertainty cloud our ability to discern trends in one critical life-history stage by observing another is poorly studied. We develop a stochastic simulation model for threatened California coho salmon Oncorhynchus kisutch to examine how well trends in one stage can be detected from observations of another. In particular, we use the model to examine the effect density dependence has on our ability to detect trends. We present a structural form for the transition between life-history stages that encompasses the common functional forms: density independence, Beverton–Holt compensatory density dependence and Ricker-type over-compensation. In small populations, such density dependence is often ignored. However, it may in fact be extremely important, for example if population decline was caused by a decrease in carrying capacity. Our results show that density dependence in any life-history transition significantly reduces the ability to detect trends in abundance; critical but inaccessible stages cannot generally be studied by monitoring more easily observed stages, especially if density dependence is present for any life-cycle transition.