One of the most studied problems in population ecology has been to understand the relative roles of top–down and bottom–up forces in regulating animal populations. This has also been a key issue in studies of vole population dyna mics. Vole populations exhibit a wide variation of dynamics, from seasonal fluctuations to multiannual variations or cyclicity. One of the hypotheses to explain cyclic population dynamics is predation by the specialist predators. A common counterargument against the predation hypothesis has been the lack of conclusive observations of the time delay in the predators’ numerical response. We studied the interaction between voles and their specialist small mustelid predators, the stoat Mustela erminea and the least weasel Mustela n. nivalis, by modelling their interaction to data sets that cover large areas of Finland. Vole abundance was monitored with biannual trappings and their predators with snow-tracking. Results show a high dependence of the predators on the voles, and this connection is generally tighter in weasels than in stoats. Weasel abundance is affected most strongly by the vole abundance in previous spring, 8.5– 10 months earlier, while in stoats the effect of autumn abundance of voles, 2.5–6 months earlier, was the strongest. These results, together with the observation that the weasels’ effects on voles are stronger after a time lag of 6–9.5 than 2–4.5 months, indicate the existence of a time lag in weasels’ numerical response. A time lag in the predators’ numerical response is a necessary condition for the predators to drive population cycles in its prey, and therefore our results support the specialist predation hypothesis.