Cyclic vole populations, defined as showing fairly regular 3–4 yr density fluctuations but with variable amplitudes, were monitored in boreal Sweden in spring and fall 1971–2002, starting in fall 1971. Voles were snap-trapped on permanent sampling plots at the landscape level within a 100 by 100 km study area north of Umeå. The predominating species trapped were Clethrionomys glareolus, C. rufocanus and Microtus agrestis. In addition to the 3–4 yr cycles, there was a long-term decrease in numbers and amplitude of the fluctuations, which was especially conspicuous in C. rufocanus. In this latter species there was a persistent decline of both spring and fall densities, apparently bringing the population close to extinction in the area. However, the decline of spring densities from the 1970s to the 1980s and onwards was also evident in C. glareolous and M. agrestis. The declines in numbers and amplitude were largely linked to an increased frequency and/or accentuation of winter declines, which more or less neutralized or even overrode the density increase during the reproductive season in the previous summer, especially so in the second year of the cycles. Thereby the gradual two-large-step build-up of high spring densities, very much founding the base for the very large peak densities and amplitudes in the 1970s, was successively replaced by a one-smaller-step build-up of more modest spring densities, leading to lower peak densities and amplitudes in the 1980s, 1990s and early 2000s. Understanding the causes of the increased frequency and/or severeness of winter declines appears critical to understanding the observed long-term changes in numbers. However, the underlying causes of the increase of winter declines and the decrease of densities and amplitudes are unknown, but some hypotheses are presented and discussed here. Also, some implications from the decreased vole abundance for reproduction and densities of predators on the voles, and on predators’ alternative prey species, are briefly discussed.