Competition is assumed to generate compensatory dynamics where an increase in one species is compensated by a decrease in others. Recently, using a community covariance technique, Houlahan et al. found that compensatory dynamics are only visible in 25–30% of natural communities studied. The study was based on scoring the sum of covariances of population densities. In contrast to the theory, and as a cautionary reminder to the interpretation of natural time series data, we show that negative community covariance can be absent even in strongly competitive communities and can be found present in communities without competitive interactions. Precise knowledge of various features of the underlying species’ biology and characteristics of the environmental variation is required before community covariance can be correctly interpreted as a proxy for the importance of competition or environmental forcing in driving community fluctuations. Other tools may therefore be more appropriate, e.g. explicit modelling of competition using modern time series analytical tools.