1. The methods of time averaging (Puccia & Levins 1985) are used to show that a positive correlation between predator and prey populations is neither necessary nor sufficient as evidence of density dependence. The sign of the correlation depends on which species is impacted by the environment. This is a general property of systems with negative feedback loops. Correlation patterns are useful for diagnosing the dynamics of populations and for evaluating control strategies. Multiple predators are modelled, with or without negative feedback or self damping in their own numbers, along with prey and plant resources, and the implications of environmental change are examined. Literature is reviewed to see how these models may help explain observed patterns. 2. Many authors have defined density dependence in terms of correlations between prey numbers and the rates of predation or parasitism. So defined, positive density dependence is less common than expected, which has led to much discussion about possible explanations and complicating factors, such as community interactions, time lags, and environmental disturbances. It is suggested here that the term density dependence be restricted to the biological effects of population density on the reproduction and/or survival of members of that population. 3. The impact of a natural enemy on the average population of its prey varies directly with the predation rate and inversely with its self damping. However, a predator that reduces the correlation between a pest and its host plant can reduce resource or yield loss without affecting the average pest population. 4. With self damped predators, environmental variability that acts directly on the plant resource generates a positive correlation between resource and consumer. Variation affecting a predator generates a negative correlation, so that pest populations are low when the resource is most available, favouring increases in that resource. Thus, for example, a biocontrol agent that is sensitive to environmental change may increase yields. A combination of predators which reduce pest numbers and also the correlation with the resource may be preferable, including predators which are not self damped but are sensitive to the environment.