1. Tanytarsus gracilentus population dynamics in Lake Myvatn show a tendency to cycle, with three oscillations occurring between 1977 and 1999 having periods of roughly 7 years. The population abundance fluctuated over four orders of magnitude. 2. A partial autocorrelation function (PACF) accounting for measurement error revealed a strong positive lag-1 autocorrelation and a moderate negative lag-2 partial autocorrelation. This suggests that the dynamics can be explained by a simple second-order autoregressive process. 3. We tested the alternative hypotheses that the cyclic dynamics of T gracilentus were driven by consumer-resource interactions in which T gracilentus is the consumer, or predator-prey interactions in which T gracilentus is the prey We analysed autoregressive models including both consumer-resource interactions and predator-prey interactions. 4. Wing length of T gracilentus was used as a surrogate for resource abundance and/or quality, because body size is known to fluctuate with resource. abundance and quality in dipterans. Furthermore, the wing lengths of Micropsectra lindrothi, a species ecologically similar to T gracilentus, fluctuated synchronously with T gracilentus wing lengths, thereby indicating that the shared resources of these two species were indeed cycling. Wing lengths of other chironomid species were not synchronized. 5. The predators of T gracilentus included midges in the genera Procladius and Macropelopia, and the fish Gasterosteus aculeatus (three-spined stickleback). 6. The autoregressive models supported the hypothesis that T gracilentus dynamics were driven by consumer-resource interactions, and rejected the hypothesis that the dynamics were driven by predator-prey interactions. 7. The models also revealed the consequences of consumer-resource interactions for the magnitude of fluctuations in T gracilentus abundance. Consumer-resource interactions amplified the exogenous variability affecting T gracilentus per capita population growth rates (e.g. temperature, rainfall, etc.), leading to variability in abundance more than two orders of magnitude greater than the exogenous variability.